Most Efficient Large Language Models for AI PC#

This page is regularly updated to help you identify the best-performing LLMs on the Intel® Core™ Ultra processor family and AI PCs. The current data is as of OpenVINO 2026.4, 23 September 2026.

The tables below list the key performance indicators for inference on built-in GPUs.

Topology

Precision

Input Size

1st latency (ms)

2nd latency (ms)

2nd token per sec

tiny-llama-1.1b-chat

INT4-MIXED

32

14.4

6.6

151.5151515

tiny-llama-1.1b-chat

INT4-MIXED

1024

40

6.8

147.0588235

tiny-llama-1.1b-chat

INT4-MIXED

32

14.9

6.9

144.9275362

tiny-llama-1.1b-chat

INT4-MIXED

1024

54.7

7.2

138.8888889

minicpm-1b-sft

INT4-MIXED

31

34.7

10.1

99.00990099

minicpm-1b-sft

INT4-MIXED

31

37

10.2

98.03921569

minicpm-1b-sft

INT4-MIXED

31

34.9

10.2

98.03921569

minicpm-1b-sft

INT4-MIXED

1014

75.1

10.7

93.45794393

minicpm-1b-sft

INT4-MIXED

1014

64.4

10.7

93.45794393

minicpm-1b-sft

INT4-MIXED

1014

66.6

10.8

92.59259259

tiny-llama-1.1b-chat

INT8-CW

32

16.3

10.8

92.59259259

tiny-llama-1.1b-chat

INT8-CW

1024

47.3

11.1

90.09009009

gemma-2b-it

INT4-MIXED

32

19.2

15.3

65.35947712

gemma-2b-it

INT4-MIXED

1024

91.3

15.6

64.1025641

minicpm-1b-sft

INT8-CW

31

43.6

15.7

63.69426752

gemma-2b-it

INT4-MIXED

32

21

15.8

63.29113924

gemma-2b-it

INT4-MIXED

1024

106.7

16.1

62.11180124

minicpm-1b-sft

INT8-CW

1014

72.2

16.5

60.60606061

gemma-2-2b

INT4-MIXED

33

26.9

17.2

58.13953488

gemma-2-2b

INT4-MIXED

1025

108.4

17.4

57.47126437

gemma-2-2b

INT4-MIXED

33

29.2

17.5

57.14285714

gemma-2-2b

INT4-MIXED

1025

130.1

17.9

55.86592179

tiny-llama-1.1b-chat

FP16

32

21.5

19.7

50.76142132

tiny-llama-1.1b-chat

FP16

1024

62.1

20

50

gemma-2b-it

INT8-CW

32

28.1

24.9

40.16064257

gemma-2b-it

INT8-CW

1024

111.5

25.3

39.5256917

gemma-2-2b

INT8-CW

33

32.5

26.2

38.16793893

gemma-2-2b

INT8-CW

1025

125.3

26.9

37.17472119

minicpm-1b-sft

FP16

31

35.4

27.9

35.84229391

minicpm-1b-sft

FP16

1014

87.8

28.9

34.60207612

codellama-7b

INT4-MIXED

32

38.2

32.5

30.76923077

codellama-7b

INT4-MIXED

32

42.1

34

29.41176471

codellama-7b

INT4-MIXED

1024

228.2

34.3

29.15451895

neural-chat-7b-v3-3

INT4-MIXED

32

41

34.6

28.9017341

zephyr-7b-beta

INT4-MIXED

32

41.1

34.6

28.9017341

zephyr-7b-beta

INT4-MIXED

1024

239.6

35.8

27.93296089

codellama-7b

INT4-MIXED

1024

280.7

35.9

27.8551532

neural-chat-7b-v3-3

INT4-MIXED

1024

239.3

35.9

27.8551532

neural-chat-7b-v3-3

INT4-MIXED

32

45.2

36.2

27.62430939

qwen-7b-chat

INT4-MIXED

32

43.1

37.1

26.9541779

neural-chat-7b-v3-3

INT4-MIXED

1024

299.9

37.5

26.66666667

qwen-7b-chat

INT4-MIXED

32

45.4

38.2

26.17801047

bloomz-7b1

INT4-MIXED

32

45.6

38.6

25.90673575

qwen-7b-chat

INT4-MIXED

1024

244.6

38.9

25.70694087

bloomz-7b1

INT4-MIXED

32

49.2

40

25

qwen-7b-chat

INT4-MIXED

1024

256.7

40

25

bloomz-7b1

INT4-MIXED

1024

325.8

40.4

24.75247525

bloomz-7b1

INT4-MIXED

1024

386.1

41.7

23.98081535

zephyr-7b-beta

INT4-MIXED

32

50.7

41.8

23.92344498

zephyr-7b-beta

INT4-MIXED

1024

258

43

23.25581395

baichuan2-7b-chat

INT4-MIXED

32

51.5

43.1

23.20185615

baichuan2-7b-chat

INT4-MIXED

1024

317.4

45

22.22222222

gemma-2b-it

FP16

32

49.4

46.5

21.50537634

gemma-2b-it

FP16

1024

166.1

46.7

21.41327623

qwen-7b-chat

INT4-MIXED

32

56.4

49.1

20.36659878

gemma-2-2b

FP16

33

53.1

50.2

19.92031873

qwen-7b-chat

INT4-MIXED

1024

266.3

51

19.60784314

gemma-2-2b

FP16

1025

162

51.1

19.56947162

dolly-v2-12b

INT4-MIXED

32

73

59.3

16.86340641

lcm-sdxl

INT8-CW

1024

62

61.4

16.28664495

codellama-7b

INT8-CW

32

67.4

61.5

16.2601626

lcm-sdxl

INT8-CW

32

62

61.5

16.2601626

dolly-v2-12b

INT4-MIXED

1024

656.3

61.9

16.15508885

codellama-7b

INT8-CW

1024

267.4

64.3

15.55209953

lcm-sdxl

INT8-HYBRID

32

65.7

64.9

15.40832049

baichuan2-7b-chat

INT8-CW

32

71.6

65

15.38461538

lcm-sdxl

INT8-HYBRID

1024

65.9

65.1

15.3609831

neural-chat-7b-v3-3

INT8-CW

32

71.8

65.5

15.26717557

zephyr-7b-beta

INT8-CW

32

72

65.5

15.26717557

qwen-7b-chat

INT8-CW

32

72

66

15.15151515

neural-chat-7b-v3-3

INT8-CW

1024

284.9

66.8

14.97005988

zephyr-7b-beta

INT8-CW

1024

283.7

66.8

14.97005988

lcm-sdxl

INT8-CW

32

67.9

67.5

14.81481481

baichuan2-7b-chat

INT8-CW

1024

305.4

67.7

14.77104874

lcm-sdxl

INT8-CW

1024

67.8

67.7

14.77104874

bloomz-7b1

INT8-CW

32

73.4

68.2

14.6627566

qwen-7b-chat

INT8-CW

1024

286.9

68.8

14.53488372

bloomz-7b1

INT8-CW

1024

389.2

70.9

14.10437236

starcoder

INT4-MIXED

32

89.2

76.2

13.12335958

starcoder

INT4-MIXED

1024

701.8

78.5

12.7388535

lcm-sdxl

FP16

32

87.6

88

11.36363636

lcm-sdxl

FP16

1024

87.5

92.1

10.8577633

Topology

Precision

Input Size

1st latency (ms)

2nd latency (ms)

2nd token per sec

distil-large-v2

INT8-CW

prompt0

226.1

9.1

109.8901099

distil-large-v2

INT8-CW

prompt1

283.9

9.9

101.010101

minicpm4-0.5b

INT4-MIXED

32

32.7

10.4

96.15384615

minicpm4-0.5b

INT4-MIXED

1024

63.7

10.6

94.33962264

minicpm4-0.5b

INT4-MIXED

32

42.6

11

90.90909091

minicpm4-0.5b

INT4-MIXED

32

37.6

11

90.90909091

whisper-large-v3-turbo

INT8-CW

prompt1

338.1

11

90.90909091

distil-large-v2

INT4-MIXED

prompt1

342.2

11.2

89.28571429

minicpm4-0.5b

INT8-CW

32

36.7

11.2

89.28571429

minicpm4-0.5b

INT4-MIXED

1024

72.8

11.3

88.49557522

distil-large-v2

INT4-MIXED

prompt0

254.6

11.4

87.71929825

minicpm4-0.5b

INT4-MIXED

1024

69.6

11.4

87.71929825

whisper-large-v3-turbo

INT4-MIXED

prompt1

364

11.5

86.95652174

minicpm4-0.5b

INT8-CW

1024

69.1

11.6

86.20689655

nuextract-tiny

INT8-CW

32

36.9

11.7

85.47008547

whisper-large-v3-turbo

INT8-CW

prompt0

266.8

11.7

85.47008547

nuextract-tiny

INT4-MIXED

32

46.8

11.9

84.03361345

nuextract-tiny

INT4-MIXED

1024

68.7

12.3

81.30081301

nuextract-tiny

INT8-CW

1024

67

12.3

81.30081301

whisper-large-v3-turbo

INT4-MIXED

prompt0

277.6

12.4

80.64516129

gemma-3-270m

INT4-MIXED

32

39.1

12.7

78.74015748

nanollava

INT4-MIXED

760

127.1

12.7

78.74015748

gemma-3-270m

INT4-MIXED

1024

51.4

12.8

78.125

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

40

12.9

77.51937984

lfm2.5-350m

INT4-MIXED

32

49.6

13

76.92307692

nanollava

INT4-MIXED

1752

170.8

13

76.92307692

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

60.3

13.2

75.75757576

nanollava

INT8-CW

760

112.3

13.4

74.62686567

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

38.2

13.4

74.62686567

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

69.7

13.5

74.07407407

nanollava

INT8-CW

1752

167.3

13.6

73.52941176

lfm2.5-350m

INT4-MIXED

1024

87.5

13.7

72.99270073

gemma-3-270m

INT8-CW

32

40

13.8

72.46376812

gemma-3-270m

INT8-CW

1024

53.8

13.9

71.94244604

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

76.1

13.9

71.94244604

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

48.6

13.9

71.94244604

whisper-small

INT4-MIXED

prompt0

124.8

14.1

70.92198582

whisper-small

INT4-MIXED

prompt0

140.5

14.2

70.42253521

lfm2.5-350m

INT8-CW

32

47.2

15

66.66666667

whisper-small

INT4-MIXED

prompt1

207.4

15

66.66666667

whisper-small

INT8-CW

prompt1

193.1

15

66.66666667

whisper-small

INT4-MIXED

prompt1

200.1

15.2

65.78947368

whisper-small

INT4-MIXED

prompt1

188.6

15.3

65.35947712

whisper-small

INT8-CW

prompt0

136

15.3

65.35947712

lfm2.5-350m

INT8-CW

1024

91.4

15.8

63.29113924

gemma-3-270m

FP16

32

41.5

15.9

62.89308176

minicpm5-1b

INT4-MIXED

32

55.9

16.1

62.11180124

tiny-llama-1.1b-chat

INT4-MIXED

32

68.1

16.1

62.11180124

whisper-large-v3-turbo

FP16

prompt0

321.9

16.1

62.11180124

gemma-3-270m

FP16

1024

53.3

16.2

61.72839506

distil-large-v2

FP16

prompt0

294.1

16.5

60.60606061

distil-large-v2

FP16

prompt1

396.1

16.5

60.60606061

whisper-large-v3-turbo

FP16

prompt1

429.9

16.5

60.60606061

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

62.5

16.7

59.88023952

tiny-llama-1.1b-chat

INT4-MIXED

1024

144

16.9

59.17159763

whisper-small

INT4-MIXED

prompt0

143.9

16.9

59.17159763

minicpm5-1b

INT4-MIXED

1024

115.6

17.1

58.47953216

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

127.8

17.2

58.13953488

llama-3.2-1b-instruct

INT4-MIXED

32

42

17.5

57.14285714

minicpm5-1b

INT8-CW

32

41.7

17.5

57.14285714

tiny-llama-1.1b-chat

INT4-MIXED

32

42.4

17.5

57.14285714

tiny-llama-1.1b-chat

INT4-MIXED

1024

116.2

18.1

55.24861878

qwen2.5-coder-0.5b-instruct

INT8-CW

32

54

18.2

54.94505495

qwen2.5-coder-0.5b-instruct

INT8-CW

1024

77.2

18.3

54.64480874

qwen3.5-0.8b

INT4-MIXED

85

104.3

18.3

54.64480874

llama-3.2-1b-instruct

INT4-MIXED

1024

130.2

18.4

54.34782609

whisper-small

FP16

prompt1

244.9

18.4

54.34782609

minicpm5-1b

INT8-CW

1024

108.5

18.5

54.05405405

qwen3.5-0.8b

INT4-MIXED

85

108.6

18.5

54.05405405

qwen3.5-0.8b

INT4-MIXED

1024

220.8

18.7

53.47593583

tiny-llama-1.1b-chat

INT8-CW

32

51.9

18.9

52.91005291

whisper-small

FP16

prompt0

141.3

18.9

52.91005291

qwen3.5-0.8b

INT4-MIXED

85

98

19

52.63157895

qwen3.5-0.8b

INT4-MIXED

1024

222.7

19.4

51.54639175

tiny-llama-1.1b-chat

INT8-CW

1024

117.6

19.6

51.02040816

qwen3.5-0.8b

INT4-MIXED

1024

237.3

19.7

50.76142132

qwen3.5-0.8b

INT8-CW

85

114.4

20

50

qwen3.5-0.8b

INT8-CW

1024

235.1

20.6

48.54368932

llama-3.2-1b-instruct

INT4-MIXED

32

42.4

21

47.61904762

exaone-4.0-1.2b

INT8-CW

32

55.5

21.1

47.39336493

qwen3-asr-0.6b

INT4-MIXED

prompt0

203.7

21.5

46.51162791

exaone-4.0-1.2b

INT4-MIXED

32

78.3

21.6

46.2962963

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

75.1

21.7

46.08294931

llama-3.2-1b-instruct

INT8-CW

32

48.5

21.8

45.87155963

llama-3.2-1b-instruct

INT4-MIXED

1024

113

21.9

45.66210046

gemma-3-1b-it

INT4-MIXED

1024

117.3

22

45.45454545

gemma-3-1b-it

INT4-MIXED

32

68.3

22

45.45454545

lfm2.5-1.2b-instruct

INT4-MIXED

32

65.3

22

45.45454545

exaone-4.0-1.2b

INT8-CW

1024

178.5

22.1

45.24886878

lfm2-1.2b

INT4-MIXED

32

62.8

22.3

44.84304933

qwen3-asr-0.6b

INT4-MIXED

prompt0

180.9

22.3

44.84304933

exaone-4.0-1.2b

INT4-MIXED

1024

191.8

22.6

44.24778761

lfm2.5-350m

FP16

32

69.1

22.6

44.24778761

llama-3.2-1b-instruct

INT8-CW

1024

124.1

22.6

44.24778761

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

147.5

22.6

44.24778761

gemma-3-1b-it

INT4-MIXED

32

68.2

23.1

43.29004329

gemma-3-1b-it

INT4-MIXED

32

58.5

23.1

43.29004329

minicpm4-0.5b

FP16

32

50.5

23.1

43.29004329

lfm2.5-1.2b-instruct

INT4-MIXED

1024

169.3

23.2

43.10344828

lfm2-1.2b

INT4-MIXED

1024

169.4

23.4

42.73504274

qwen3-asr-0.6b

INT4-MIXED

prompt1

492.4

23.5

42.55319149

qwen3-asr-0.6b

INT4-MIXED

prompt1

485.8

23.6

42.37288136

gemma-3-1b-it

INT4-MIXED

1024

123

23.7

42.19409283

gemma-3-1b-it

INT4-MIXED

1024

140.6

23.9

41.84100418

lfm2.5-350m

FP16

1024

99.5

24

41.66666667

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

81.4

24.1

41.49377593

qwen3-asr-0.6b

INT4-MIXED

prompt0

196.6

24.2

41.32231405

exaone-4.0-1.2b

INT4-MIXED

32

88.9

24.3

41.15226337

minicpm4-0.5b

FP16

1024

89.8

24.3

41.15226337

lfm2-1.2b

INT4-MIXED

32

76.4

24.6

40.6504065

lfm2.5-1.2b-instruct

INT4-MIXED

32

77.8

24.6

40.6504065

qwen3-asr-0.6b

INT8-CW

prompt0

188.3

24.6

40.6504065

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

182.9

24.7

40.48582996

exaone-4.0-1.2b

INT4-MIXED

1024

220.9

25

40

nuextract-tiny

FP16

32

81.3

25

40

qwen2.5-1.5b-instruct

INT4-MIXED

32

88.7

25

40

exaone-4.0-1.2b

INT4-MIXED

32

69.6

25.3

39.5256917

qwen3-asr-0.6b

INT8-CW

prompt1

467.6

25.3

39.5256917

qwen2.5-coder-0.5b-instruct

FP16

32

67.3

25.4

39.37007874

qwen3-asr-0.6b

INT4-MIXED

prompt1

511.8

25.4

39.37007874

qwen2.5-1.5b-instruct

INT4-MIXED

1024

184.3

25.5

39.21568627

lfm2-1.2b

INT4-MIXED

1024

196.5

25.6

39.0625

minicpm-1b-sft

INT4-MIXED

31

98.1

25.6

39.0625

nuextract-tiny

FP16

1024

85.3

25.6

39.0625

qwen2.5-1.5b-instruct

INT4-MIXED

32

67.5

25.7

38.91050584

qwen2.5-coder-0.5b-instruct

FP16

1024

90.5

25.7

38.91050584

lfm2.5-1.2b-instruct

INT4-MIXED

1024

200.7

25.8

38.75968992

exaone-4.0-1.2b

INT4-MIXED

1024

194.5

26.1

38.31417625

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

87.7

26.2

38.16793893

minicpm-1b-sft

INT4-MIXED

31

92.3

26.4

37.87878788

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

69.1

26.4

37.87878788

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

156.8

26.5

37.73584906

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

194.7

26.8

37.31343284

minicpm-1b-sft

INT4-MIXED

1014

169.5

26.8

37.31343284

nanollava

FP16

760

176.8

26.8

37.31343284

qwen2.5-1.5b-instruct

INT4-MIXED

1024

165

26.9

37.17472119

lfm2-1.2b

INT4-MIXED

32

55.2

27

37.03703704

lfm2.5-1.2b-instruct

INT4-MIXED

32

67.8

27.1

36.900369

deepseek-r1-distill-qwen-1.5b

INT8-CW

32

75.9

27.3

36.63003663

lfm2.5-1.2b-instruct

INT4-MIXED

1024

176.6

27.3

36.63003663

minicpm-1b-sft

INT4-MIXED

31

95.5

27.4

36.49635036

lfm2-1.2b

INT4-MIXED

1024

168.7

27.5

36.36363636

nanollava

FP16

1752

225.7

27.5

36.36363636

qwen2.5-1.5b-instruct

INT8-CW

32

70.5

27.8

35.97122302

qwen2.5-coder-1.5b-instruct

INT8-CW

32

65.8

27.8

35.97122302

gemma-2b-it

INT4-MIXED

32

67.4

27.9

35.84229391

lfm2-1.2b

INT4-MIXED

32

86

28

35.71428571

deepseek-r1-distill-qwen-1.5b

INT8-CW

1024

164.3

28.1

35.58718861

lfm2.5-1.2b-instruct

INT4-MIXED

32

60.7

28.1

35.58718861

qwen2.5-1.5b-instruct

INT8-CW

1024

164

28.2

35.46099291

qwen2.5-coder-1.5b-instruct

INT8-CW

1024

164.4

28.3

35.33568905

gemma-3-1b-it

INT8-CW

32

88.7

28.4

35.21126761

gemma-2b-it

INT4-MIXED

1024

214.2

28.6

34.96503497

lfm2-1.2b

INT8-CW

32

77

28.7

34.84320557

gemma-3-1b-it

INT8-CW

1024

143.9

29

34.48275862

lfm2.5-1.2b-instruct

INT4-MIXED

1024

186

29

34.48275862

minicpm-1b-sft

INT8-CW

31

95

29.2

34.24657534

lfm2-1.2b

INT4-MIXED

1024

183.6

29.3

34.12969283

gemma-2b-it

INT4-MIXED

1024

269.1

29.5

33.89830508

lfm2.5-1.2b-instruct

INT8-CW

32

69.6

29.5

33.89830508

minicpm-1b-sft

INT4-MIXED

1014

201.6

29.5

33.89830508

minicpm-1b-sft

INT4-MIXED

1014

174.9

29.6

33.78378378

gemma-2b-it

INT4-MIXED

32

73.6

29.7

33.67003367

lfm2.5-1.2b-instruct

INT8-CW

1024

199.4

30.1

33.22259136

minicpm-1b-sft

INT8-CW

1014

182.7

30.4

32.89473684

lfm2-1.2b

INT8-CW

1024

198.3

30.7

32.5732899

lfm2-8b-a1b

INT4-MIXED

32

166.7

32.2

31.05590062

lfm2-8b-a1b

INT4-MIXED

1024

1150.3

33.4

29.94011976

gemma-2-2b

INT4-MIXED

33

82.8

34.8

28.73563218

phi-2

INT4-MIXED

32

140.5

34.9

28.65329513

smollm3-3b

INT4-MIXED

32

107

35

28.57142857

gemma-4-e2b-it

INT4-MIXED

274

427.5

35.2

28.40909091

gemma-2-2b

INT4-MIXED

1025

262.4

35.4

28.24858757

gemma-4-e2b-it

INT4-MIXED

1024

592

36.1

27.70083102

smollm3-3b

INT4-MIXED

1024

289.3

36.1

27.70083102

llama-3.2-3b-instruct

INT4-MIXED

32

71.6

36.3

27.54820937

qwen2.5-coder-3b-instruct

INT4-MIXED

32

106.3

36.4

27.47252747

gemma-2-2b

INT4-MIXED

33

90.5

36.6

27.32240437

minicpm5-1b

FP16

32

77.3

36.7

27.2479564

stable-zephyr-3b-dpo

INT4-MIXED

32

130.1

36.7

27.2479564

stablelm-3b-4e1t

INT4-MIXED

32

115.6

36.8

27.17391304

phi-2

INT4-MIXED

1024

328.1

37

27.02702703

afm-4.5b

INT4-MIXED

32

83.2

37.1

26.9541779

llama-3.2-3b-instruct

INT4-MIXED

1024

275.6

37.2

26.88172043

smollm3-3b

INT4-MIXED

32

123.4

37.2

26.88172043

gemma-2-2b

INT4-MIXED

1025

313.4

37.4

26.73796791

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

272.4

37.6

26.59574468

tiny-llama-1.1b-chat

FP16

32

54.7

37.7

26.52519894

afm-4.5b

INT4-MIXED

1024

431.8

38

26.31578947

minicpm5-1b

FP16

1024

151.1

38

26.31578947

phi-2

INT4-MIXED

32

129.4

38.1

26.24671916

llama-3.2-3b-instruct

INT4-MIXED

32

114.7

38.2

26.17801047

qwen3.5-0.8b

FP16

85

149.2

38.8

25.77319588

smollm3-3b

INT4-MIXED

1024

377.5

38.8

25.77319588

stable-zephyr-3b-dpo

INT4-MIXED

1024

339.8

39

25.64102564

stablelm-3b-4e1t

INT4-MIXED

1024

333.1

39

25.64102564

qwen2.5-coder-3b-instruct

INT4-MIXED

32

132.5

39.1

25.57544757

gemma-4-e2b-it

INT4-MIXED

274

454.1

39.3

25.44529262

tiny-llama-1.1b-chat

FP16

1024

177.2

39.4

25.38071066

phi-3-mini-4k-instruct

INT4-MIXED

32

111.7

39.6

25.25252525

granite-4.0-h-micro

INT4-MIXED

33

182.3

39.7

25.18891688

phi-3-mini-128k-instruct

INT4-MIXED

32

119.6

39.7

25.18891688

phi-3.5-mini-instruct

INT4-MIXED

32

129.8

39.9

25.06265664

qwen3.5-0.8b

FP16

1024

296

39.9

25.06265664

stable-zephyr-3b-dpo

INT4-MIXED

32

145.6

39.9

25.06265664

llama-3.2-3b-instruct

INT4-MIXED

1024

381.7

40

25

gemma-4-e2b-it

INT4-MIXED

1024

661

40.1

24.93765586

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

414.1

40.2

24.87562189

granite-4.0-h-micro

INT4-MIXED

1033

767.8

40.5

24.69135802

gemma-2b-it

INT8-CW

32

59.6

40.6

24.63054187

phi-2

INT4-MIXED

1024

422.1

41

24.3902439

qwen3-asr-0.6b

FP16

prompt1

592.5

41.3

24.21307506

gemma-2b-it

INT8-CW

1024

237.1

41.4

24.15458937

lfm2-8b-a1b

INT8-CW

32

191

41.4

24.15458937

llama-3.2-3b-instruct

INT4-MIXED

32

92

41.8

23.92344498

phi-3-mini-128k-instruct

INT4-MIXED

1024

444.9

41.8

23.92344498

phi-3-mini-4k-instruct

INT4-MIXED

1024

442.2

41.8

23.92344498

phi-3.5-mini-instruct

INT4-MIXED

1024

441.7

41.9

23.86634845

gemma-3-1b-it

FP16

32

75.9

42

23.80952381

phi-3-mini-128k-instruct

INT4-MIXED

32

119.6

42

23.80952381

stable-zephyr-3b-dpo

INT4-MIXED

1024

407.4

42.2

23.69668246

granite-4.0-h-micro

INT4-MIXED

33

185.7

42.4

23.58490566

gemma-3-1b-it

FP16

1024

174.8

42.5

23.52941176

gemma-4-e2b-it

INT8-CW

274

430.3

42.8

23.36448598

lfm2-8b-a1b

INT8-CW

1024

1083.3

42.8

23.36448598

llama-3.2-3b-instruct

INT4-MIXED

1024

294.5

42.8

23.36448598

granite-4.0-h-micro

INT4-MIXED

1033

867.7

42.9

23.31002331

qwen3-asr-0.6b

FP16

prompt0

247

42.9

23.31002331

phi-3.5-vision-instruct

INT4-MIXED

802

320.9

43.3

23.09468822

gemma-4-e2b-it

INT8-CW

1024

607.3

43.6

22.93577982

gemma-2-2b

INT8-CW

33

77.5

44

22.72727273

phi-3-mini-128k-instruct

INT4-MIXED

1024

596.2

44

22.72727273

phi-3.5-vision-instruct

INT4-MIXED

1032

614.8

44

22.72727273

phi-2

INT8-CW

32

126.6

45.1

22.172949

phi-3-mini-4k-instruct

INT4-MIXED

32

111.1

45.2

22.12389381

stablelm-3b-4e1t

INT4-MIXED

32

116.3

45.4

22.02643172

gemma-2-2b

INT8-CW

1025

277.4

45.6

21.92982456

llama-3.2-1b-instruct

FP16

32

66.9

45.6

21.92982456

stablelm-3b-4e1t

INT8-CW

32

139

45.8

21.83406114

phi-3.5-mini-instruct

INT4-MIXED

32

118.7

46

21.73913043

qwen3-4b

INT4-MIXED

32

135.9

46.1

21.69197397

stable-zephyr-3b-dpo

INT8-CW

32

116.8

46.2

21.64502165

phi-4-mini-reasoning

INT4-MIXED

32

148.8

46.4

21.55172414

phi-4-mini-instruct

INT4-MIXED

32

153.3

46.6

21.45922747

llama-3.2-1b-instruct

FP16

1024

195.6

46.7

21.41327623

phi-3-mini-4k-instruct

INT4-MIXED

1024

585.9

47.8

20.92050209

phi-4-mini-instruct

INT4-MIXED

1024

388.5

48

20.83333333

phi-4-mini-reasoning

INT4-MIXED

1024

388.1

48

20.83333333

qwen3-4b

INT4-MIXED

1024

413.3

48.2

20.74688797

phi-2

INT8-CW

1024

352

48.4

20.66115702

stablelm-3b-4e1t

INT4-MIXED

1024

359.3

48.4

20.66115702

phi-4-mini-reasoning

INT4-MIXED

32

159.3

48.5

20.6185567

phi-3.5-mini-instruct

INT4-MIXED

1024

610.6

48.7

20.5338809

gemma-3-4b-it

INT4-MIXED

32

152.4

48.8

20.49180328

stablelm-3b-4e1t

INT8-CW

1024

364.4

49.2

20.32520325

stable-zephyr-3b-dpo

INT8-CW

1024

363

49.3

20.28397566

gemma-3-4b-it

INT4-MIXED

1024

421.1

49.8

20.08032129

exaone-4.0-1.2b

FP16

32

107

50

20

llama-3.2-3b-instruct

INT8-CW

32

89

50

20

phi-4-mini-reasoning

INT4-MIXED

1024

495.2

50.3

19.88071571

qwen2.5-coder-3b-instruct

INT8-CW

32

107.1

50.4

19.84126984

smollm3-3b

INT8-CW

32

95.3

50.5

19.8019802

phi-3-mini-4k-instruct

INT4-MIXED

32

115.6

50.7

19.72386588

phi-3.5-mini-instruct

INT4-MIXED

32

114.9

51

19.60784314

afm-4.5b

INT8-CW

32

70.8

51.1

19.56947162

exaone-4.0-1.2b

FP16

1024

266.7

51.3

19.49317739

gemma-3-4b-it

INT4-MIXED

32

150.3

51.7

19.34235977

llama-3.2-3b-instruct

INT8-CW

1024

316.9

51.7

19.34235977

qwen3-vl-4b-instruct

INT4-MIXED

4937

1278.7

51.8

19.30501931

qwen3-vl-4b-instruct

INT4-MIXED

4907

1248.4

51.9

19.26782274

lfm2-1.2b

FP16

32

92.5

52

19.23076923

lfm2.5-1.2b-instruct

FP16

32

92.7

52.1

19.19385797

lfm2.5-1.2b-instruct

FP16

1024

276.7

52.2

19.15708812

qwen2.5-coder-3b-instruct

INT8-CW

1024

389.8

52.2

19.15708812

lfm2-1.2b

FP16

1024

267.9

52.3

19.12045889

gemma-3-4b-it

INT4-MIXED

1024

530.5

52.4

19.08396947

phi-4-mini-instruct

INT4-MIXED

32

160.6

52.4

19.08396947

smollm3-3b

INT8-CW

1024

318.8

52.6

19.01140684

afm-4.5b

INT8-CW

1024

325.2

52.7

18.97533207

phi-3-mini-4k-instruct

INT4-MIXED

1024

445.6

53.2

18.79699248

phi-3.5-mini-instruct

INT4-MIXED

1024

451.7

53.4

18.72659176

phi-4-mini-instruct

INT4-MIXED

1024

539

53.5

18.69158879

qwen2.5-coder-3b-instruct

INT4-MIXED

32

58.1

53.5

18.69158879

smollm3-3b

INT4-MIXED

32

66.8

53.8

18.58736059

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

288.9

54.3

18.41620626

smollm3-3b

INT4-MIXED

1024

309.9

54.6

18.31501832

qwen3-vl-4b-instruct

INT4-MIXED

4937

1283.2

54.8

18.24817518

minicpm-1b-sft

FP16

31

131.7

55

18.18181818

deepseek-r1-distill-qwen-1.5b

FP16

32

101.1

55.3

18.08318264

phi-4-mini-instruct

INT4-MIXED

32

145.5

55.3

18.08318264

qwen3-vl-4b-instruct

INT4-MIXED

4907

1248.6

55.3

18.08318264

granite-4.0-h-micro

INT8-CW

33

185.8

55.7

17.95332136

deepseek-r1-distill-qwen-1.5b

FP16

1024

239.1

55.8

17.92114695

phi-4-mini-reasoning

INT4-MIXED

32

133.1

56.1

17.82531194

granite-4.0-h-micro

INT8-CW

1033

792.2

56.5

17.69911504

minicpm-1b-sft

FP16

1014

254.4

57

17.54385965

qwen2.5-3b-instruct-gptq

INT4-MIXED

32

77.8

57.2

17.48251748

phi-4-mini-instruct

INT4-MIXED

1024

404.5

57.4

17.42160279

phi-4-mini-reasoning

INT4-MIXED

1024

407

57.4

17.42160279

granite-4.0-h-micro

INT4-MIXED

33

110.9

57.5

17.39130435

qwen2.5-coder-1.5b-instruct

FP16

32

91.1

57.5

17.39130435

whisper-large-v3

INT4-MIXED

prompt1

549.4

57.5

17.39130435

qwen2.5-1.5b-instruct

FP16

32

93.5

57.6

17.36111111

granite-4.0-h-micro

INT4-MIXED

1033

743.8

57.8

17.30103806

whisper-large-v3

INT8-CW

prompt1

539.4

57.9

17.27115717

whisper-large-v3

INT4-MIXED

prompt0

489.8

58

17.24137931

gemma-3n-e2b-it

INT4-MIXED

275

710.9

58.1

17.21170396

codellama-7b

INT4-MIXED

32

114.7

58.2

17.18213058

gemma-3n-e2b-it

INT4-MIXED

1025

1074.1

58.2

17.18213058

phi-3-mini-4k-instruct

INT8-CW

32

98.5

58.3

17.15265866

qwen2.5-1.5b-instruct

FP16

1024

241.1

58.3

17.15265866

llama-2-7b-chat-hf

INT4-MIXED

32

103.3

58.5

17.09401709

qwen2.5-3b-instruct-gptq

INT4-MIXED

1024

307.9

58.5

17.09401709

qwen2.5-coder-1.5b-instruct

FP16

1024

235.2

58.5

17.09401709

phi-3-mini-128k-instruct

INT8-CW

32

97.8

58.6

17.06484642

phi-3.5-mini-instruct

INT8-CW

32

99.9

58.7

17.03577513

whisper-large-v3

INT8-CW

prompt0

468.6

58.7

17.03577513

qwen3-4b

INT4-MIXED

32

116.1

60.2

16.61129568

gemma-3-4b-it

INT4-MIXED

32

128.7

61.3

16.3132137

biomistral-7b-slerp

INT4-MIXED

7

92.4

61.4

16.28664495

codellama-7b

INT4-MIXED

32

105.4

61.4

16.28664495

codellama-7b

INT4-MIXED

1024

617

61.5

16.2601626

falcon-7b-instruct

INT4-MIXED

32

125

61.5

16.2601626

gemma-3-4b-it

INT4-MIXED

1024

443.4

61.5

16.2601626

zephyr-7b-beta

INT4-MIXED

32

112.9

61.6

16.23376623

phi-4-multimodal-instruct

INT4-MIXED

1362

1277.5

61.7

16.20745543

neural-chat-7b-v3-3

INT4-MIXED

32

102.2

61.8

16.18122977

phi-4-multimodal-instruct

INT4-MIXED

578

529.9

61.8

16.18122977

phi-3.5-vision-instruct

INT8-CW

802

322.4

61.9

16.15508885

lfm2-24b-a2b

INT4-MIXED

32

459.8

62.1

16.10305958

phi-4-multimodal-instruct

INT4-MIXED

786

492.5

62.1

16.10305958

phi-3-mini-4k-instruct

INT8-CW

1024

490.8

62.2

16.07717042

phi-4-multimodal-instruct

INT4-MIXED

1570

860.1

62.2

16.07717042

llama-2-7b-chat-hf

INT4-MIXED

1024

627

62.3

16.05136437

mistral-7b-instruct-v0.3

INT4-MIXED

32

101.4

62.4

16.02564103

phi-3.5-mini-instruct

INT8-CW

1024

495

62.4

16.02564103

phi-4-mini-instruct

INT8-CW

32

128.1

62.4

16.02564103

mistral-7b-instruct-v0.2

INT4-MIXED

32

106.3

62.5

16

phi-4-mini-reasoning

INT8-CW

32

127.1

62.5

16

phi-3-mini-128k-instruct

INT8-CW

1024

493.8

62.6

15.97444089

phi-3.5-vision-instruct

INT8-CW

1032

555.1

62.8

15.92356688

qwen3-4b

INT4-MIXED

1024

434

63

15.87301587

neural-chat-7b-v3-3

INT4-MIXED

1024

648.2

63.3

15.79778831

zephyr-7b-beta

INT4-MIXED

1024

661.2

63.3

15.79778831

minicpm3-4b

INT4-MIXED

32

263.1

63.6

15.72327044

falcon-7b-instruct

INT4-MIXED

1024

658.2

63.8

15.67398119

gemma-3-4b-it

INT8-CW

32

136

64.1

15.60062402

mistral-7b-instruct-v0.2

INT4-MIXED

1024

651.3

64.1

15.60062402

phi-4-mini-reasoning

INT8-CW

1024

446

64.1

15.60062402

qwen3-4b

INT8-CW

32

112.2

64.1

15.60062402

mistral-7b-instruct-v0.3

INT4-MIXED

1024

660.6

64.2

15.57632399

phi-4-mini-instruct

INT8-CW

1024

439.3

64.3

15.55209953

qwen2.5-7b-instruct

INT4-MIXED

32

89.9

64.4

15.52795031

qwen2.5-7b-instruct-1m

INT4-MIXED

32

107.3

64.4

15.52795031

gemma-3n-e2b-it

INT4-MIXED

275

785.8

64.5

15.50387597

biomistral-7b-slerp

INT4-MIXED

7

83.8

64.6

15.47987616

minicpm-o-2_6

INT4-MIXED

238

653.1

64.7

15.45595054

codellama-7b

INT4-MIXED

1024

882.3

64.8

15.43209877

minicpm-v-2_6

INT4-MIXED

228

657.9

64.9

15.40832049

neural-chat-7b-v3-3

INT4-MIXED

32

114

65

15.38461538

mistral-7b-instruct-v0.2

INT4-MIXED

32

111.3

65.1

15.3609831

mistral-7b-instruct-v0.3

INT4-MIXED

32

115.5

65.1

15.3609831

mistral-7b-instruct-v0.1

INT4-MIXED

32

128.9

65.2

15.33742331

qwen2.5-vl-7b-instruct

INT4-MIXED

32

141.2

65.4

15.29051988

gemma-3-4b-it

INT8-CW

1024

460.9

65.8

15.19756839

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

607.8

65.9

15.17450683

bloomz-7b1

INT4-MIXED

32

104.5

66

15.15151515

qwen2.5-7b-instruct

INT4-MIXED

1024

606.3

66.1

15.12859304

gemma-3n-e2b-it

INT8-CW

275

731.1

66.2

15.10574018

qwen3-vl-4b-instruct

INT4-MIXED

4907

1264.4

66.2

15.10574018

qwen-7b-chat

INT4-MIXED

32

127.6

66.4

15.06024096

neural-chat-7b-v3-3

INT4-MIXED

1024

911.7

66.5

15.03759398

qwen3-vl-4b-instruct

INT4-MIXED

4937

1289.5

66.6

15.01501502

whisper-large-v3

FP16

prompt1

600

66.6

15.01501502

deepseek-r1-distill-llama-8b

INT4-MIXED

35

118.3

66.9

14.94768311

llama-3-8b-instruct

INT4-MIXED

32

112.4

66.9

14.94768311

mistral-7b-instruct-v0.3

INT4-MIXED

1024

941.1

67

14.92537313

llama-3.1-8b-instruct

INT4-MIXED

32

106.4

67.3

14.85884101

mistral-7b-instruct-v0.1

INT4-MIXED

1025

921.1

67.3

14.85884101

qwen3-4b

INT8-CW

1024

454.4

67.3

14.85884101

gemma-3n-e2b-it

INT4-MIXED

1025

1154.7

67.4

14.83679525

mistral-7b-instruct-v0.2

INT4-MIXED

1024

941.8

67.4

14.83679525

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

670.2

67.5

14.81481481

gpt-oss-20b

INT4-MIXED

32

252.6

67.8

14.74926254

gemma-3n-e2b-it

INT8-CW

1025

1128.6

68.2

14.6627566

qwen2.5-7b-instruct

INT4-MIXED

32

101.3

68.5

14.59854015

gpt-oss-20b

INT4-MIXED

1024

3036.6

68.6

14.57725948

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

110.7

68.7

14.55604076

minicpm3-4b

INT4-MIXED

1024

711.4

68.9

14.5137881

qwen2-7b-instruct

INT4-MIXED

32

113.2

69.1

14.47178003

qwen2.5-7b-instruct-1m

INT4-MIXED

32

111.1

69.2

14.45086705

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

744.2

69.3

14.43001443

llama-3-8b-instruct

INT4-MIXED

1024

649.6

69.3

14.43001443

llama-3.1-8b-instruct

INT4-MIXED

1024

657.9

69.3

14.43001443

minicpm-o-2_6

INT4-MIXED

238

662

69.3

14.43001443

bloomz-7b1

INT4-MIXED

1024

837.1

69.4

14.4092219

minicpm-v-2_6

INT4-MIXED

228

636.9

69.5

14.38848921

minicpm3-4b

INT4-MIXED

32

286.6

69.5

14.38848921

minicpm4-8b

INT4-MIXED

32

127.4

69.5

14.38848921

qwen3-8b

INT4-MIXED

32

137.6

69.5

14.38848921

bloomz-7b1

INT4-MIXED

32

117.7

69.7

14.3472023

qwen-7b-chat

INT4-MIXED

1024

631.3

69.9

14.30615165

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

833.2

70

14.28571429

qwen2.5-7b-instruct

INT4-MIXED

1024

853.3

70

14.28571429

whisper-large-v3

FP16

prompt0

538.2

70.1

14.26533524

fara-7b

INT4-MIXED

32

235.7

70.2

14.24501425

qwen2.5-vl-7b-instruct

INT4-MIXED

32

141.4

70.3

14.22475107

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

71.8

70.3

14.22475107

qwen2-7b-instruct

INT4-MIXED

1024

851.4

70.4

14.20454545

minicpm-v-4_5

INT4-MIXED

217

682.2

70.5

14.18439716

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

854.3

70.5

14.18439716

minicpm4-8b

INT4-MIXED

1024

739

71.3

14.02524544

qwen3-8b

INT4-MIXED

1024

689.6

71.3

14.02524544

llama-3-8b-instruct

INT4-MIXED

32

121.4

71.4

14.00560224

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

891.8

71.5

13.98601399

falcon-7b-instruct

INT4-MIXED

32

129.6

71.6

13.96648045

fara-7b

INT4-MIXED

1024

1080.6

71.7

13.94700139

llava-next-video-7b-hf

INT4-MIXED

2945

593.9

72.1

13.86962552

minicpm3-4b

INT4-MIXED

32

231.6

72.1

13.86962552

minicpm4-8b

INT4-MIXED

32

112.2

72.7

13.75515818

bloomz-7b1

INT4-MIXED

1024

1079.7

72.8

13.73626374

llama-3-8b-instruct

INT4-MIXED

1024

915.9

72.8

13.73626374

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

76.4

72.8

13.73626374

qwen3-vl-4b-instruct

INT8-CW

4907

1277.1

72.9

13.71742112

qwen3-vl-4b-instruct

INT8-CW

4937

1331.3

73

13.69863014

falcon-7b-instruct

INT4-MIXED

1024

919.5

73.3

13.6425648

qwen3-8b

INT4-MIXED

32

149.4

74.2

13.47708895

minicpm4-8b

INT4-MIXED

1024

1011.6

74.5

13.42281879

minicpm-v-4_5

INT4-MIXED

217

707.5

74.6

13.40482574

minicpm3-4b

INT8-CW

32

257.7

74.7

13.38688086

baichuan2-7b-chat

INT4-MIXED

32

135.1

74.8

13.36898396

minicpm3-4b

INT4-MIXED

1024

850.3

75.6

13.22751323

qwen3_8b_draft_eagle3

INT4-MIXED

32

110.6

75.9

13.17523057

qwen3-8b

INT4-MIXED

1024

959

76.1

13.14060447

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

108.8

76.6

13.05483029

phi-4-multimodal-instruct

INT4-MIXED

786

526.6

77.8

12.85347044

minicpm3-4b

INT4-MIXED

1024

748.1

78

12.82051282

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

614.1

78.2

12.78772379

phi-4-multimodal-instruct

INT4-MIXED

578

685.6

78.2

12.78772379

baichuan2-7b-chat

INT4-MIXED

1024

878.4

78.6

12.72264631

videochat-flash-qwen2_5-7b_internvideo2-1b

INT4-MIXED

14531

60873.7

78.6

12.72264631

deepseek-r1-distill-llama-8b

INT4-MIXED

35

126.1

78.7

12.7064803

qwen3_8b_draft_eagle3

INT4-MIXED

1024

909.1

79.3

12.61034048

phi-4-multimodal-instruct

INT4-MIXED

1570

909

79.4

12.59445844

qwen3.5-9b

INT4-MIXED

83

325.3

79.4

12.59445844

gemma-7b-it

INT4-MIXED

32

112.9

79.5

12.57861635

phi-4-multimodal-instruct

INT4-MIXED

1362

1610.4

79.5

12.57861635

qwen3.5-9b

INT4-MIXED

1024

1120.3

79.5

12.57861635

phi-4-multimodal-instruct

INT8-CW

578

576.9

80.1

12.48439451

phi-4-multimodal-instruct

INT8-CW

786

510.6

80.1

12.48439451

gpt-oss-20b

INT4-MIXED

32

283.4

80.2

12.46882793

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

769.1

80.3

12.45330012

phi-4-multimodal-instruct

INT8-CW

1362

1370.9

81.4

12.28501229

phi-4-multimodal-instruct

INT8-CW

1570

857.6

81.5

12.26993865

gemma-7b-it

INT4-MIXED

1024

769.2

81.6

12.25490196

llama-2-7b-chat-hf

INT4-MIXED

32

97.3

82.1

12.18026797

qwen3.5-9b

INT4-MIXED

83

356.9

82.3

12.15066829

minicpm3-4b

INT8-CW

1024

786.4

82.5

12.12121212

deepseek-r1-distill-llama-8b

INT4-MIXED

35

128.5

82.8

12.07729469

gemma-3n-e4b-it

INT4-MIXED

275

868.6

82.8

12.07729469

gemma-2b-it

FP16

32

91.7

83.1

12.03369434

gemma-3n-e4b-it

INT4-MIXED

1025

1410

83.1

12.03369434

qwen3.5-9b

INT4-MIXED

1024

1369.3

83.2

12.01923077

gemma-2b-it

FP16

1024

391.7

83.4

11.99040767

gpt-oss-20b

INT4-MIXED

1024

3255.1

83.5

11.9760479

llama-3.1-8b-instruct

INT4-MIXED

32

134.3

83.5

11.9760479

gemma-7b-it

INT4-MIXED

32

116.8

83.7

11.9474313

qwen2-7b-instruct

INT4-MIXED

32

105.2

83.7

11.9474313

qwen2-7b-instruct-gptq

INT4-MIXED

32

99.3

83.8

11.93317422

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

888.3

84

11.9047619

qwen2.5-7b-instruct

INT4-MIXED

32

103.1

84.1

11.89060642

minicpm-v-2_6

INT4-MIXED

228

603.4

84.5

11.83431953

mistral-7b-instruct-v0.2

INT4-MIXED

32

98.3

84.5

11.83431953

mistral-7b-instruct-v0.3

INT4-MIXED

32

107.1

84.6

11.82033097

llama-2-7b-chat-hf

INT4-MIXED

1024

628.9

84.8

11.79245283

llama-3.1-8b-instruct

INT4-MIXED

1024

889.2

84.9

11.77856302

qwen-7b-chat

INT4-MIXED

32

122.3

84.9

11.77856302

qwen2-7b-instruct

INT4-MIXED

1024

603.2

85.1

11.75088132

qwen2.5-vl-7b-instruct

INT4-MIXED

32

143.9

85.2

11.7370892

gemma-3n-e4b-it

INT4-MIXED

275

858.8

85.4

11.70960187

qwen2-7b-instruct-gptq

INT4-MIXED

1024

612.6

85.4

11.70960187

zephyr-7b-beta

INT4-MIXED

32

114.2

85.4

11.70960187

qwen2.5-7b-instruct

INT4-MIXED

1024

626.2

85.7

11.66861144

gemma-7b-it

INT4-MIXED

1024

1085.2

86.4

11.57407407

mistral-7b-instruct-v0.2

INT4-MIXED

1024

649.7

86.5

11.56069364

gemma-3n-e4b-it

INT4-MIXED

1025

1512.7

86.6

11.54734411

mistral-7b-instruct-v0.3

INT4-MIXED

1024

660.8

86.7

11.53402537

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

672.9

87.1

11.48105626

zephyr-7b-beta

INT4-MIXED

1024

662.2

87.4

11.4416476

gemma-2-2b

FP16

33

99.1

88.1

11.3507378

qwen-7b-chat

INT4-MIXED

1024

664.8

88.1

11.3507378

gemma-2-9b-it

INT4-MIXED

32

180.5

88.2

11.33786848

gemma-2-2b

FP16

1025

429.8

89.7

11.14827202

llama-3-8b-instruct

INT4-MIXED

32

113.5

90.2

11.0864745

llama-3-8b-instruct

INT4-MIXED

32

112.4

90.4

11.0619469

llama-3.1-8b-instruct

INT4-MIXED

32

103.7

90.5

11.04972376

phi-2

FP16

32

127.5

91.1

10.97694841

gemma-2-9b-it

INT4-MIXED

1024

889.5

91.2

10.96491228

llama-3-8b-instruct

INT4-MIXED

1024

674.8

92

10.86956522

llama-3-8b-instruct

INT4-MIXED

1024

659.4

92.4

10.82251082

stable-zephyr-3b-dpo

FP16

32

134.3

92.6

10.79913607

stablelm-3b-4e1t

FP16

32

154.2

92.6

10.79913607

llama-3.1-8b-instruct

INT4-MIXED

1024

650.9

92.7

10.78748652

qwen3-8b

INT4-MIXED

32

131.7

93

10.75268817

qwen-7b-chat

INT4-MIXED

32

129.9

93.8

10.66098081

qwen3-8b

INT4-MIXED

1024

717.3

95.1

10.51524711

phi-2

FP16

1024

553.5

95.7

10.44932079

qwen-7b-chat

INT4-MIXED

1024

659.8

97.1

10.29866117

stable-zephyr-3b-dpo

FP16

1024

563.7

97.4

10.26694045

gemma-3n-e2b-it

FP16

275

696.9

97.6

10.24590164

stablelm-3b-4e1t

FP16

1024

562.5

97.8

10.22494888

gemma-3n-e2b-it

FP16

1025

1243.1

98.4

10.16260163

gemma-3n-e4b-it

INT8-CW

275

854.9

98.5

10.15228426

llama-2-7b-chat-hf

INT8-CW

32

130.4

98.7

10.13171226

gemma-3n-e4b-it

INT8-CW

1025

1456

99.5

10.05025126

Topology

Precision

Input Size

1st latency (ms)

2nd latency (ms)

2nd token per sec

lfm2.5-350m

INT4-MIXED

1024

39.9

3.6

277.7777778

lfm2.5-350m

INT4-MIXED

32

12.9

3.6

277.7777778

minicpm4-0.5b

INT4-MIXED

32

17.9

4.7

212.7659574

gemma-3-270m

INT4-MIXED

1024

26.5

4.8

208.3333333

gemma-3-270m

INT4-MIXED

32

18.6

4.8

208.3333333

lfm2.5-350m

INT8-CW

32

13.8

4.9

204.0816327

lfm2.5-350m

INT8-CW

1024

37.7

5.1

196.0784314

minicpm4-0.5b

INT4-MIXED

32

23

5.1

196.0784314

minicpm4-0.5b

INT4-MIXED

32

19.7

5.1

196.0784314

gemma-3-270m

INT8-CW

32

20

5.2

192.3076923

gemma-3-270m

INT8-CW

1024

25.7

5.3

188.6792453

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

18.4

5.4

185.1851852

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

18.7

5.4

185.1851852

minicpm4-0.5b

INT4-MIXED

1024

39.7

5.5

181.8181818

distil-large-v2

INT4-MIXED

prompt0

158.6

5.6

178.5714286

whisper-large-v3-turbo

INT4-MIXED

prompt0

166.9

5.7

175.4385965

distil-large-v2

INT4-MIXED

prompt1

213.7

5.8

172.4137931

distil-large-v2

INT8-CW

prompt0

155

5.8

172.4137931

distil-large-v2

INT8-CW

prompt1

206.9

5.8

172.4137931

whisper-large-v3-turbo

INT4-MIXED

prompt1

224.1

5.8

172.4137931

minicpm4-0.5b

INT4-MIXED

1024

41.9

5.9

169.4915254

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

23.1

5.9

169.4915254

whisper-large-v3-turbo

INT8-CW

prompt1

223.5

5.9

169.4915254

minicpm4-0.5b

INT4-MIXED

1024

48.8

6

166.6666667

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

42.1

6.1

163.9344262

whisper-large-v3-turbo

INT8-CW

prompt0

161.8

6.1

163.9344262

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

43.4

6.2

161.2903226

nuextract-tiny

INT4-MIXED

32

19.7

6.3

158.7301587

whisper-small

INT4-MIXED

prompt0

78.8

6.4

156.25

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

51

6.6

151.5151515

whisper-small

INT4-MIXED

prompt0

77.9

6.6

151.5151515

whisper-small

INT4-MIXED

prompt1

136.7

6.7

149.2537313

nuextract-tiny

INT4-MIXED

1024

48.6

6.8

147.0588235

whisper-small

INT4-MIXED

prompt1

134.2

6.8

147.0588235

whisper-small

INT8-CW

prompt0

81.2

6.9

144.9275362

nanollava

INT4-MIXED

760

84.3

7

142.8571429

whisper-small

INT8-CW

prompt1

138.1

7

142.8571429

minicpm4-0.5b

INT8-CW

32

24.7

7.2

138.8888889

nanollava

INT4-MIXED

1752

129.5

7.3

136.9863014

nuextract-tiny

INT8-CW

32

25.5

7.3

136.9863014

whisper-small

INT4-MIXED

prompt0

81.9

7.3

136.9863014

whisper-small

INT4-MIXED

prompt1

134.8

7.3

136.9863014

qwen2.5-coder-0.5b-instruct

INT8-CW

32

26.2

7.6

131.5789474

gemma-3-270m

FP16

32

16.6

7.7

129.8701299

minicpm4-0.5b

INT8-CW

1024

44.2

7.7

129.8701299

whisper-large-v3-turbo

FP16

prompt0

167.5

7.9

126.5822785

whisper-large-v3-turbo

FP16

prompt1

221.8

7.9

126.5822785

distil-large-v2

FP16

prompt0

160.7

8

125

distil-large-v2

FP16

prompt1

220.8

8

125

gemma-3-270m

FP16

1024

28.2

8

125

nanollava

INT8-CW

760

90

8

125

nuextract-tiny

INT8-CW

1024

44.8

8

125

qwen2.5-coder-0.5b-instruct

INT8-CW

1024

47.4

8

125

tiny-llama-1.1b-chat

INT4-MIXED

32

26.1

8.1

123.4567901

lfm2.5-350m

FP16

32

13.3

8.2

121.9512195

lfm2.5-350m

FP16

1024

43.6

8.4

119.047619

minicpm5-1b

INT4-MIXED

32

29.2

8.5

117.6470588

lfm2-1.2b

INT4-MIXED

32

20.9

8.6

116.2790698

lfm2.5-1.2b-instruct

INT4-MIXED

32

19.1

8.6

116.2790698

nanollava

INT8-CW

1752

127.2

8.7

114.9425287

tiny-llama-1.1b-chat

INT4-MIXED

32

26.6

8.7

114.9425287

lfm2-1.2b

INT4-MIXED

32

22

8.8

113.6363636

lfm2.5-1.2b-instruct

INT4-MIXED

1024

92.5

8.8

113.6363636

whisper-small

FP16

prompt0

76.2

8.8

113.6363636

lfm2-1.2b

INT4-MIXED

1024

94.7

8.9

112.3595506

lfm2.5-1.2b-instruct

INT4-MIXED

32

22.2

8.9

112.3595506

tiny-llama-1.1b-chat

INT4-MIXED

1024

92.7

8.9

112.3595506

lfm2-1.2b

INT4-MIXED

1024

98

9

111.1111111

lfm2.5-1.2b-instruct

INT4-MIXED

32

22.3

9

111.1111111

whisper-small

FP16

prompt1

136.1

9

111.1111111

lfm2-1.2b

INT4-MIXED

32

22.6

9.2

108.6956522

lfm2.5-1.2b-instruct

INT4-MIXED

1024

109.1

9.2

108.6956522

lfm2.5-1.2b-instruct

INT4-MIXED

1024

98.1

9.2

108.6956522

lfm2-1.2b

INT4-MIXED

1024

109.3

9.3

107.5268817

minicpm5-1b

INT4-MIXED

1024

81.6

9.3

107.5268817

tiny-llama-1.1b-chat

INT4-MIXED

1024

111.7

9.4

106.3829787

qwen3.5-0.8b

INT4-MIXED

85

55.5

9.9

101.010101

qwen3.5-0.8b

INT4-MIXED

85

57.8

10

100

llama-3.2-1b-instruct

INT4-MIXED

32

22.3

10.1

99.00990099

qwen3-asr-0.6b

INT4-MIXED

prompt0

95.5

10.1

99.00990099

llama-3.2-1b-instruct

INT4-MIXED

32

22.4

10.2

98.03921569

qwen3-asr-0.6b

INT4-MIXED

prompt0

99.2

10.2

98.03921569

qwen3.5-0.8b

INT4-MIXED

85

61.2

10.2

98.03921569

qwen3-asr-0.6b

INT4-MIXED

prompt1

279.6

10.5

95.23809524

qwen3-asr-0.6b

INT8-CW

prompt0

99.7

10.5

95.23809524

qwen3.5-0.8b

INT4-MIXED

1024

133.8

10.5

95.23809524

qwen3-asr-0.6b

INT4-MIXED

prompt1

293.8

10.6

94.33962264

qwen3.5-0.8b

INT4-MIXED

1024

141.8

10.6

94.33962264

gemma-3-1b-it

INT4-MIXED

32

35

10.7

93.45794393

qwen3.5-0.8b

INT4-MIXED

1024

149

10.7

93.45794393

gemma-3-1b-it

INT4-MIXED

32

36.5

10.8

92.59259259

llama-3.2-1b-instruct

INT4-MIXED

1024

93

10.8

92.59259259

gemma-3-1b-it

INT4-MIXED

32

34.6

10.9

91.74311927

gemma-3-1b-it

INT4-MIXED

1024

83.7

10.9

91.74311927

exaone-4.0-1.2b

INT4-MIXED

32

42.6

11

90.90909091

llama-3.2-1b-instruct

INT4-MIXED

1024

103.3

11

90.90909091

gemma-3-1b-it

INT4-MIXED

1024

87.5

11.1

90.09009009

minicpm4-0.5b

FP16

32

23.3

11.1

90.09009009

qwen3-asr-0.6b

INT8-CW

prompt1

288.1

11.1

90.09009009

exaone-4.0-1.2b

INT4-MIXED

32

42.4

11.4

87.71929825

gemma-3-1b-it

INT4-MIXED

1024

93.6

11.4

87.71929825

exaone-4.0-1.2b

INT4-MIXED

32

44.6

11.5

86.95652174

minicpm4-0.5b

FP16

1024

58.3

11.5

86.95652174

nuextract-tiny

FP16

32

26.7

11.5

86.95652174

minicpm5-1b

INT8-CW

32

31.3

11.6

86.20689655

qwen3.5-0.8b

INT8-CW

85

72.9

11.6

86.20689655

qwen3-asr-0.6b

INT4-MIXED

prompt0

135.2

11.9

84.03361345

exaone-4.0-1.2b

INT4-MIXED

1024

143.4

12.1

82.6446281

qwen3.5-0.8b

INT8-CW

1024

151.6

12.1

82.6446281

nanollava

FP16

760

91.7

12.2

81.96721311

qwen2.5-coder-0.5b-instruct

FP16

32

24.3

12.3

81.30081301

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

33.3

12.3

81.30081301

qwen3-asr-0.6b

INT4-MIXED

prompt1

297.7

12.3

81.30081301

exaone-4.0-1.2b

INT4-MIXED

1024

148.4

12.4

80.64516129

nuextract-tiny

FP16

1024

55.4

12.4

80.64516129

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

33.3

12.5

80

minicpm5-1b

INT8-CW

1024

82.6

12.5

80

qwen2.5-1.5b-instruct

INT4-MIXED

32

35.3

12.5

80

qwen2.5-coder-0.5b-instruct

FP16

1024

53.3

12.5

80

exaone-4.0-1.2b

INT4-MIXED

1024

156.1

12.6

79.36507937

lfm2-8b-a1b

INT4-MIXED

32

163.3

12.6

79.36507937

qwen2.5-1.5b-instruct

INT4-MIXED

32

33.8

12.6

79.36507937

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

37.7

12.6

79.36507937

tiny-llama-1.1b-chat

INT8-CW

32

29.4

12.6

79.36507937

lfm2-8b-a1b

INT4-MIXED

1024

346.2

12.7

78.74015748

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

34.1

12.7

78.74015748

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

122.3

13

76.92307692

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

152.6

13.2

75.75757576

nanollava

FP16

1752

142.7

13.3

75.18796992

qwen2.5-1.5b-instruct

INT4-MIXED

1024

153.1

13.3

75.18796992

qwen2.5-1.5b-instruct

INT4-MIXED

1024

153.4

13.3

75.18796992

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

160.3

13.3

75.18796992

qwen2.5-1.5b-instruct

INT4-MIXED

32

36.6

13.4

74.62686567

qwen2.5-1.5b-instruct

INT4-MIXED

32

39.3

13.4

74.62686567

tiny-llama-1.1b-chat

INT8-CW

1024

106.4

13.4

74.62686567

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

153.6

13.6

73.52941176

minicpm-1b-sft

INT4-MIXED

31

53.4

13.7

72.99270073

gemma-3-1b-it

INT8-CW

32

38.2

13.9

71.94244604

lfm2.5-1.2b-instruct

INT4-MIXED

32

20.4

13.9

71.94244604

lfm2-1.2b

INT4-MIXED

32

20.6

14

71.42857143

minicpm-1b-sft

INT4-MIXED

31

55

14

71.42857143

lfm2.5-1.2b-instruct

INT8-CW

32

23.1

14.1

70.92198582

qwen2.5-1.5b-instruct

INT4-MIXED

1024

159.8

14.1

70.92198582

qwen2.5-1.5b-instruct

INT4-MIXED

1024

158.7

14.1

70.92198582

gemma-3-1b-it

INT8-CW

1024

93.6

14.2

70.42253521

lfm2-1.2b

INT8-CW

32

27.4

14.2

70.42253521

lfm2.5-1.2b-instruct

INT4-MIXED

1024

104.1

14.2

70.42253521

lfm2-1.2b

INT4-MIXED

1024

105.3

14.3

69.93006993

minicpm-1b-sft

INT4-MIXED

31

56.9

14.3

69.93006993

lfm2.5-1.2b-instruct

INT8-CW

1024

114.9

14.4

69.44444444

lfm2-1.2b

INT8-CW

1024

117.4

14.5

68.96551724

minicpm-1b-sft

INT4-MIXED

1014

119.8

14.5

68.96551724

minicpm-1b-sft

INT4-MIXED

1014

129.1

14.9

67.11409396

llama-3.2-1b-instruct

INT8-CW

32

24

15.1

66.22516556

minicpm-1b-sft

INT4-MIXED

1014

143.4

15.1

66.22516556

qwen3-asr-0.6b

FP16

prompt0

104.5

15.3

65.35947712

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

36.3

15.9

62.89308176

llama-3.2-1b-instruct

INT8-CW

1024

104.9

15.9

62.89308176

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

160.8

16.5

60.60606061

qwen3-asr-0.6b

FP16

prompt1

307.4

16.6

60.24096386

exaone-4.0-1.2b

INT8-CW

32

46.9

17.2

58.13953488

qwen2.5-1.5b-instruct

INT4-MIXED

32

63

18

55.55555556

gemma-2b-it

INT4-MIXED

32

27.9

18.1

55.24861878

lfm2-8b-a1b

INT8-CW

32

114.7

18.2

54.94505495

exaone-4.0-1.2b

INT8-CW

1024

152.3

18.3

54.64480874

qwen3.5-0.8b

FP16

85

58.9

18.4

54.34782609

lfm2-8b-a1b

INT8-CW

1024

400.6

18.6

53.76344086

deepseek-r1-distill-qwen-1.5b

INT8-CW

32

38.6

18.7

53.47593583

qwen2.5-1.5b-instruct

INT4-MIXED

32

65.7

18.7

53.47593583

qwen3.5-0.8b

FP16

1024

169.9

18.7

53.47593583

gemma-2b-it

INT4-MIXED

1024

173

18.8

53.19148936

qwen2.5-1.5b-instruct

INT4-MIXED

1024

186.8

18.8

53.19148936

qwen2.5-1.5b-instruct

INT8-CW

32

38.9

18.8

53.19148936

qwen2.5-1.5b-instruct

INT8-CW

32

40.3

18.8

53.19148936

qwen2.5-coder-1.5b-instruct

INT8-CW

32

39.4

18.8

53.19148936

lfm2-24b-a2b

INT4-MIXED

32

242.4

18.9

52.91005291

gemma-2b-it

INT4-MIXED

32

30.5

19

52.63157895

phi-2

INT4-MIXED

32

52.5

19.1

52.35602094

lfm2-24b-a2b

INT4-MIXED

1024

776.8

19.2

52.08333333

minicpm-1b-sft

INT8-CW

31

66.4

19.3

51.8134715

stablelm-3b-4e1t

INT4-MIXED

32

56.3

19.3

51.8134715

deepseek-r1-distill-qwen-1.5b

INT8-CW

1024

139.1

19.4

51.54639175

qwen2.5-1.5b-instruct

INT4-MIXED

1024

192.6

19.4

51.54639175

stable-zephyr-3b-dpo

INT4-MIXED

32

58.5

19.5

51.28205128

gemma-4-e2b-it

INT4-MIXED

274

289

19.6

51.02040816

qwen2.5-1.5b-instruct

INT8-CW

1024

138.2

19.6

51.02040816

qwen2.5-1.5b-instruct

INT8-CW

1024

139.5

19.6

51.02040816

qwen2.5-coder-1.5b-instruct

INT8-CW

1024

138.5

19.6

51.02040816

gemma-2b-it

INT4-MIXED

1024

211.9

19.8

50.50505051

gemma-4-e2b-it

INT4-MIXED

1024

463.7

20

50

whisper-large-v3

INT4-MIXED

prompt0

237.7

20.2

49.5049505

whisper-large-v3

INT4-MIXED

prompt1

298.5

20.3

49.26108374

gemma-2-2b

INT4-MIXED

33

39.3

20.5

48.7804878

gemma-4-e2b-it

INT4-MIXED

274

309.4

20.6

48.54368932

smollm3-3b

INT4-MIXED

32

45.9

20.7

48.30917874

minicpm-1b-sft

INT8-CW

1014

143.8

20.8

48.07692308

phi-2

INT4-MIXED

32

56.8

20.8

48.07692308

minicpm5-1b

FP16

32

27.1

20.9

47.84688995

gemma-4-e2b-it

INT4-MIXED

1024

492.1

21

47.61904762

gemma-2-2b

INT4-MIXED

33

40.3

21.2

47.16981132

gemma-2-2b

INT4-MIXED

1025

204.3

21.5

46.51162791

phi-2

INT4-MIXED

1024

289.2

21.5

46.51162791

smollm3-3b

INT4-MIXED

32

48.5

21.6

46.2962963

stable-zephyr-3b-dpo

INT4-MIXED

1024

288.6

21.7

46.08294931

stablelm-3b-4e1t

INT4-MIXED

1024

287.5

21.7

46.08294931

qwen2.5-coder-3b-instruct

INT4-MIXED

32

47.8

21.8

45.87155963

minicpm5-1b

FP16

1024

94.9

21.9

45.66210046

qwen2.5-coder-3b-instruct

INT4-MIXED

32

46.8

21.9

45.66210046

smollm3-3b

INT4-MIXED

32

47

21.9

45.66210046

stable-zephyr-3b-dpo

INT4-MIXED

32

64.6

21.9

45.66210046

smollm3-3b

INT4-MIXED

1024

228.3

22

45.45454545

gemma-2-2b

INT4-MIXED

1025

229.7

22.4

44.64285714

llama-3.2-3b-instruct

INT4-MIXED

32

38.8

22.4

44.64285714

smollm3-3b

INT4-MIXED

1024

243.9

22.9

43.66812227

granite-4.0-h-micro

INT4-MIXED

33

89.8

23.1

43.29004329

llama-3.2-3b-instruct

INT4-MIXED

32

37.2

23.2

43.10344828

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

223.1

23.2

43.10344828

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

238.8

23.2

43.10344828

smollm3-3b

INT4-MIXED

1024

275.5

23.2

43.10344828

granite-4.0-h-micro

INT4-MIXED

1033

692.5

23.3

42.91845494

qwen2.5-coder-3b-instruct

INT4-MIXED

32

52.3

23.3

42.91845494

llama-3.2-3b-instruct

INT4-MIXED

32

39.4

23.4

42.73504274

phi-2

INT4-MIXED

1024

361.7

23.5

42.55319149

stablelm-3b-4e1t

INT4-MIXED

32

61.5

23.5

42.55319149

gemma-3-1b-it

FP16

32

35.3

23.6

42.37288136

gemma-3-1b-it

FP16

1024

107

23.8

42.01680672

whisper-large-v3

INT8-CW

prompt0

237.9

23.8

42.01680672

llama-3.2-3b-instruct

INT4-MIXED

1024

234.6

23.9

41.84100418

granite-4.0-h-micro

INT4-MIXED

33

94.7

24

41.66666667

granite-4.0-h-micro

INT4-MIXED

1033

711.6

24.1

41.49377593

phi-3-mini-128k-instruct

INT4-MIXED

32

43.4

24.1

41.49377593

phi-3-mini-4k-instruct

INT4-MIXED

32

43.2

24.2

41.32231405

phi-3.5-mini-instruct

INT4-MIXED

32

42

24.2

41.32231405

qwen2.5-1.5b-instruct

INT8-CW

32

70.1

24.2

41.32231405

stable-zephyr-3b-dpo

INT4-MIXED

1024

325.4

24.2

41.32231405

granite-4.0-h-micro

INT4-MIXED

33

97

24.3

41.15226337

granite-4.0-h-micro

INT4-MIXED

1033

766

24.4

40.98360656

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

372.6

24.6

40.6504065

whisper-large-v3

INT8-CW

prompt1

299.7

24.6

40.6504065

llama-3.2-3b-instruct

INT4-MIXED

1024

251.8

24.8

40.32258065

llama-3.2-3b-instruct

INT4-MIXED

1024

273

25.1

39.84063745

qwen2.5-1.5b-instruct

INT8-CW

1024

179.6

25.1

39.84063745

phi-3-mini-4k-instruct

INT4-MIXED

32

45.2

25.2

39.68253968

phi-3.5-mini-instruct

INT4-MIXED

32

44.2

25.3

39.5256917

qwen2.5-3b-instruct-gptq

INT4-MIXED

32

51.5

25.6

39.0625

stablelm-3b-4e1t

INT4-MIXED

1024

316.8

25.6

39.0625

phi-3-mini-128k-instruct

INT4-MIXED

32

44.9

25.8

38.75968992

tiny-llama-1.1b-chat

FP16

32

29.3

26.2

38.16793893

gemma-3n-e2b-it

INT4-MIXED

275

528.8

26.3

38.02281369

gemma-3n-e2b-it

INT4-MIXED

1025

892

26.5

37.73584906

phi-3-mini-128k-instruct

INT4-MIXED

1024

368.4

26.7

37.45318352

phi-3-mini-4k-instruct

INT4-MIXED

1024

366.9

26.7

37.45318352

phi-3.5-mini-instruct

INT4-MIXED

1024

366.4

26.8

37.31343284

qwen2.5-3b-instruct-gptq

INT4-MIXED

1024

240.8

26.9

37.17472119

gemma-3n-e2b-it

INT4-MIXED

275

530.5

27

37.03703704

gemma-3n-e2b-it

INT4-MIXED

1025

919.7

27.1

36.900369

tiny-llama-1.1b-chat

FP16

1024

115.7

27.1

36.900369

phi-3-mini-4k-instruct

INT4-MIXED

32

43.7

27.2

36.76470588

qwen3-4b

INT4-MIXED

32

53.8

27.3

36.63003663

phi-3.5-mini-instruct

INT4-MIXED

32

49

27.4

36.49635036

phi-3-mini-4k-instruct

INT4-MIXED

1024

390.1

27.7

36.10108303

phi-3.5-vision-instruct

INT4-MIXED

802

308.4

27.7

36.10108303

phi-3.5-mini-instruct

INT4-MIXED

1024

390.7

27.8

35.97122302

phi-3-mini-128k-instruct

INT4-MIXED

1024

488.6

28.2

35.46099291

phi-3.5-vision-instruct

INT4-MIXED

1032

802.4

28.2

35.46099291

gemma-3-4b-it

INT4-MIXED

32

62.1

28.3

35.33568905

phi-4-mini-instruct

INT4-MIXED

32

57.9

28.3

35.33568905

phi-4-mini-reasoning

INT4-MIXED

32

58.1

28.4

35.21126761

gpt-oss-20b

INT4-MIXED

32

239.5

28.5

35.0877193

qwen3-4b

INT4-MIXED

32

56

28.5

35.0877193

lfm2-1.2b

FP16

32

33.6

28.8

34.72222222

lfm2.5-1.2b-instruct

FP16

32

33.7

28.9

34.60207612

afm-4.5b

INT4-MIXED

32

47.9

29

34.48275862

lfm2-1.2b

FP16

1024

143.1

29

34.48275862

lfm2.5-1.2b-instruct

FP16

1024

142

29.1

34.36426117

phi-4-mini-instruct

INT4-MIXED

32

60

29.2

34.24657534

phi-4-mini-reasoning

INT4-MIXED

32

58.4

29.2

34.24657534

gemma-2b-it

INT8-CW

32

37.2

29.3

34.12969283

gemma-3-4b-it

INT4-MIXED

1024

335.3

29.3

34.12969283

gemma-3-4b-it

INT4-MIXED

32

63.5

29.3

34.12969283

gpt-oss-20b

INT4-MIXED

1024

679.5

29.3

34.12969283

phi-4-mini-reasoning

INT4-MIXED

32

59.9

29.4

34.01360544

phi-3-mini-4k-instruct

INT4-MIXED

1024

515.6

29.7

33.67003367

phi-3.5-mini-instruct

INT4-MIXED

1024

557.3

29.9

33.44481605

qwen3-4b

INT4-MIXED

1024

337.3

29.9

33.44481605

phi-4-mini-instruct

INT4-MIXED

1024

368

30

33.33333333

phi-4-mini-reasoning

INT4-MIXED

1024

369.8

30

33.33333333

gemma-4-e2b-it

INT8-CW

274

315.2

30.1

33.22259136

gemma-2b-it

INT8-CW

1024

210.4

30.2

33.11258278

gemma-3-4b-it

INT4-MIXED

1024

356.6

30.3

33.00330033

llama-3.2-1b-instruct

FP16

32

34.9

30.4

32.89473684

afm-4.5b

INT4-MIXED

1024

391.1

30.6

32.67973856

gemma-4-e2b-it

INT8-CW

1024

504.9

30.8

32.46753247

phi-4-mini-instruct

INT4-MIXED

1024

390

30.9

32.36245955

phi-4-mini-reasoning

INT4-MIXED

1024

392.1

31

32.25806452

qwen3-4b

INT4-MIXED

1024

355.5

31

32.25806452

llama-3.2-1b-instruct

FP16

1024

134.9

31.1

32.15434084

phi-4-mini-instruct

INT4-MIXED

32

63.8

31.2

32.05128205

gemma-2-2b

INT8-CW

33

46

31.3

31.94888179

phi-4-mini-reasoning

INT4-MIXED

1024

411.2

31.3

31.94888179

qwen3-vl-4b-instruct

INT4-MIXED

4907

1073.2

31.8

31.44654088

qwen3-vl-4b-instruct

INT4-MIXED

4937

1135

31.8

31.44654088

minicpm3-4b

INT4-MIXED

32

122.2

32

31.25

phi-2

INT8-CW

32

59.7

32.5

30.76923077

stable-zephyr-3b-dpo

INT8-CW

32

60.3

32.5

30.76923077

stablelm-3b-4e1t

INT8-CW

32

60.5

32.6

30.67484663

gemma-2-2b

INT8-CW

1025

235.1

32.8

30.48780488

phi-4-mini-instruct

INT4-MIXED

1024

475.3

32.9

30.39513678

qwen3-vl-4b-instruct

INT4-MIXED

4907

1087.2

33

30.3030303

qwen3-vl-4b-instruct

INT4-MIXED

4937

1145.6

33.1

30.21148036

minicpm-1b-sft

FP16

31

52.3

33.4

29.94011976

minicpm3-4b

INT4-MIXED

32

123.3

33.4

29.94011976

qwen3-vl-4b-instruct

INT4-MIXED

4907

1073.1

33.4

29.94011976

qwen3-vl-4b-instruct

INT4-MIXED

4937

1121.2

33.4

29.94011976

minicpm3-4b

INT4-MIXED

32

131.4

33.7

29.6735905

phi-4-multimodal-instruct

INT4-MIXED

578

447.6

34.1

29.3255132

whisper-large-v3

FP16

prompt0

231.8

34.2

29.23976608

qwen2.5-1.5b-instruct

FP16

32

38.2

34.3

29.15451895

deepseek-r1-distill-qwen-1.5b

FP16

32

38.4

34.4

29.06976744

phi-4-multimodal-instruct

INT4-MIXED

786

322.7

34.4

29.06976744

qwen2.5-coder-1.5b-instruct

FP16

32

38.8

34.4

29.06976744

smollm3-3b

INT8-CW

32

53.2

34.4

29.06976744

qwen2.5-coder-3b-instruct

INT8-CW

32

55.4

34.7

28.8184438

exaone-4.0-1.2b

FP16

32

39.3

34.8

28.73563218

stablelm-3b-4e1t

INT8-CW

1024

319.7

34.8

28.73563218

minicpm-1b-sft

FP16

1014

180.3

34.9

28.65329513

phi-2

INT8-CW

1024

319.4

34.9

28.65329513

stable-zephyr-3b-dpo

INT8-CW

1024

320.2

35.1

28.49002849

deepseek-r1-distill-qwen-1.5b

FP16

1024

169.4

35.2

28.40909091

qwen2.5-1.5b-instruct

FP16

1024

167.7

35.3

28.3286119

qwen2.5-coder-1.5b-instruct

FP16

1024

168.9

35.3

28.3286119

whisper-large-v3

FP16

prompt1

293.5

35.3

28.3286119

phi-4-multimodal-instruct

INT4-MIXED

1362

1063.8

35.5

28.16901408

phi-4-multimodal-instruct

INT4-MIXED

1570

631.7

35.6

28.08988764

gemma-3n-e2b-it

INT8-CW

275

551.5

35.7

28.01120448

smollm3-3b

INT8-CW

1024

273.7

35.9

27.8551532

exaone-4.0-1.2b

FP16

1024

178.6

36

27.77777778

gemma-3n-e2b-it

INT8-CW

1025

939.2

36.1

27.70083102

gpt-oss-20b

INT4-MIXED

32

233.3

36.1

27.70083102

qwen2.5-coder-3b-instruct

INT8-CW

1024

269.9

36.1

27.70083102

minicpm3-4b

INT4-MIXED

1024

757.3

36.2

27.62430939

gpt-oss-20b

INT4-MIXED

1024

696.9

37.1

26.9541779

minicpm3-4b

INT4-MIXED

1024

796.8

37.6

26.59574468

gemma-3n-e4b-it

INT4-MIXED

275

566.6

37.9

26.38522427

llama-3.2-3b-instruct

INT8-CW

32

46.9

38

26.31578947

minicpm3-4b

INT4-MIXED

1024

841

38

26.31578947

codellama-7b

INT4-MIXED

32

56.8

38.4

26.04166667

gemma-3n-e4b-it

INT4-MIXED

1025

1107.2

38.4

26.04166667

llama-2-7b-chat-hf

INT4-MIXED

32

56.8

38.5

25.97402597

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

39.3

38.7

25.83979328

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

40.1

39.6

25.25252525

llama-3.2-3b-instruct

INT8-CW

1024

270.9

39.7

25.18891688

gemma-3n-e4b-it

INT4-MIXED

275

569

39.8

25.12562814

gemma-3n-e4b-it

INT4-MIXED

1025

1169.3

40.2

24.87562189

phi-4-multimodal-instruct

INT4-MIXED

578

507.6

40.3

24.81389578

qwen3_8b_draft_eagle3

INT4-MIXED

32

65.6

40.3

24.81389578

granite-4.0-h-micro

INT8-CW

33

94.8

40.4

24.75247525

phi-4-multimodal-instruct

INT4-MIXED

786

357.7

40.5

24.69135802

biomistral-7b-slerp

INT4-MIXED

7

47.1

40.6

24.63054187

granite-4.0-h-micro

INT8-CW

1033

740.1

40.6

24.63054187

llama-2-7b-chat-hf

INT4-MIXED

32

60.9

40.6

24.63054187

mistral-7b-instruct-v0.2

INT4-MIXED

32

60.9

40.8

24.50980392

neural-chat-7b-v3-3

INT4-MIXED

32

60.5

40.8

24.50980392

zephyr-7b-beta

INT4-MIXED

32

61

40.8

24.50980392

mistral-7b-instruct-v0.3

INT4-MIXED

32

61.2

40.9

24.44987775

codellama-7b

INT4-MIXED

32

62.1

41

24.3902439

codellama-7b

INT4-MIXED

1024

510.3

41.2

24.27184466

llama-2-7b-chat-hf

INT4-MIXED

1024

512

41.4

24.15458937

falcon-7b-instruct

INT4-MIXED

32

57.8

41.6

24.03846154

phi-4-multimodal-instruct

INT4-MIXED

1362

1209

41.6

24.03846154

phi-4-multimodal-instruct

INT4-MIXED

1570

668.2

41.8

23.92344498

neural-chat-7b-v3-3

INT4-MIXED

1024

536.3

42.7

23.41920375

zephyr-7b-beta

INT4-MIXED

1024

535.3

42.7

23.41920375

mistral-7b-instruct-v0.2

INT4-MIXED

1024

536.1

42.8

23.36448598

mistral-7b-instruct-v0.3

INT4-MIXED

1024

535

42.9

23.31002331

qwen2.5-7b-instruct

INT4-MIXED

32

58.1

42.9

23.31002331

qwen2.5-7b-instruct

INT4-MIXED

32

59

43

23.25581395

mistral-7b-instruct-v0.3

INT4-MIXED

32

65.8

43.1

23.20185615

qwen2.5-7b-instruct-1m

INT4-MIXED

32

57.1

43.1

23.20185615

mistral-7b-instruct-v0.2

INT4-MIXED

32

65.6

43.2

23.14814815

qwen3_8b_draft_eagle3

INT4-MIXED

1024

642.1

43.2

23.14814815

qwen2.5-vl-7b-instruct

INT4-MIXED

32

67.7

43.3

23.09468822

biomistral-7b-slerp

INT4-MIXED

7

50.3

43.4

23.04147465

minicpm-o-2_6

INT4-MIXED

238

508.4

43.4

23.04147465

minicpm-v-2_6

INT4-MIXED

228

499.5

43.4

23.04147465

llama-2-7b-chat-hf

INT4-MIXED

1024

546.8

43.5

22.98850575

mistral-7b-instruct-v0.1

INT4-MIXED

32

67

43.6

22.93577982

mistral-7b-instruct-v0.2

INT4-MIXED

32

67.6

43.6

22.93577982

neural-chat-7b-v3-3

INT4-MIXED

32

66.3

43.6

22.93577982

mistral-7b-instruct-v0.3

INT4-MIXED

32

65.7

43.7

22.88329519

qwen-7b-chat

INT4-MIXED

32

63.7

43.8

22.83105023

codellama-7b

INT4-MIXED

1024

609.6

44.1

22.67573696

falcon-7b-instruct

INT4-MIXED

1024

551.7

44.1

22.67573696

phi-3-mini-4k-instruct

INT8-CW

32

58.9

44.2

22.62443439

phi-3-mini-128k-instruct

INT8-CW

32

58.2

44.3

22.57336343

phi-3.5-mini-instruct

INT8-CW

32

58.8

44.4

22.52252252

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

60.9

44.7

22.37136465

deepseek-r1-distill-llama-8b

INT4-MIXED

35

70.9

44.8

22.32142857

llama-3-8b-instruct

INT4-MIXED

32

65.7

44.8

22.32142857

llama-3.1-8b-instruct

INT4-MIXED

32

65.8

44.8

22.32142857

qwen2.5-7b-instruct

INT4-MIXED

32

61.3

44.8

22.32142857

qwen2.5-7b-instruct

INT4-MIXED

1024

485.6

44.9

22.27171492

qwen2.5-7b-instruct

INT4-MIXED

32

61

44.9

22.27171492

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

495.7

44.9

22.27171492

gemma-3-4b-it

INT8-CW

32

67.8

45

22.22222222

minicpm-v-2_6

INT4-MIXED

228

511.2

45

22.22222222

qwen2-7b-instruct

INT4-MIXED

32

61.1

45

22.22222222

qwen2.5-7b-instruct

INT4-MIXED

1024

494.5

45

22.22222222

qwen2-7b-instruct-gptq

INT4-MIXED

32

61.3

45.1

22.172949

qwen2.5-7b-instruct

INT4-MIXED

32

60.8

45.1

22.172949

qwen2.5-7b-instruct-1m

INT4-MIXED

32

61.1

45.1

22.172949

qwen2.5-vl-7b-instruct

INT4-MIXED

32

70.2

45.1

22.172949

mistral-7b-instruct-v0.2

INT4-MIXED

1024

568.6

45.2

22.12389381

mistral-7b-instruct-v0.3

INT4-MIXED

1024

568.7

45.2

22.12389381

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

513.5

45.2

22.12389381

qwen2-7b-instruct

INT4-MIXED

32

61.6

45.3

22.07505519

qwen2.5-7b-instruct

INT4-MIXED

32

61.6

45.3

22.07505519

qwen2.5-vl-7b-instruct

INT4-MIXED

32

71.2

45.3

22.07505519

fara-7b

INT4-MIXED

32

104.5

45.4

22.02643172

minicpm-v-2_6

INT4-MIXED

228

525.2

45.4

22.02643172

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

61.9

45.5

21.97802198

bloomz-7b1

INT4-MIXED

32

71.3

45.6

21.92982456

minicpm-o-2_6

INT4-MIXED

238

518.9

45.6

21.92982456

mistral-7b-instruct-v0.1

INT4-MIXED

1025

688.7

45.6

21.92982456

mistral-7b-instruct-v0.2

INT4-MIXED

1024

655

45.6

21.92982456

mistral-7b-instruct-v0.3

INT4-MIXED

1024

656.2

45.7

21.88183807

neural-chat-7b-v3-3

INT4-MIXED

1024

653.9

45.7

21.88183807

phi-4-mini-instruct

INT8-CW

32

62.3

45.9

21.78649237

minicpm-v-4_5

INT4-MIXED

217

531.2

46.1

21.69197397

phi-4-mini-reasoning

INT8-CW

32

63.3

46.2

21.64502165

falcon-7b-instruct

INT4-MIXED

32

63.7

46.3

21.59827214

qwen-7b-chat

INT4-MIXED

32

68.2

46.3

21.59827214

qwen3-8b

INT4-MIXED

32

63.4

46.3

21.59827214

minicpm4-8b

INT4-MIXED

32

68.2

46.4

21.55172414

phi-3.5-vision-instruct

INT8-CW

802

334.9

46.4

21.55172414

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

515.4

46.5

21.50537634

gemma-3-4b-it

INT8-CW

1024

377.1

46.5

21.50537634

qwen2.5-7b-instruct

INT4-MIXED

1024

527.1

46.5

21.50537634

qwen2-7b-instruct

INT4-MIXED

1024

534.1

46.6

21.45922747

qwen2-7b-instruct-gptq

INT4-MIXED

1024

535.9

46.6

21.45922747

qwen2.5-7b-instruct

INT4-MIXED

1024

537.5

46.6

21.45922747

qwen-7b-chat

INT4-MIXED

1024

540.1

46.7

21.41327623

llama-3.1-8b-instruct

INT4-MIXED

1024

538.6

46.8

21.36752137

llama-3-8b-instruct

INT4-MIXED

1024

539.4

46.9

21.32196162

phi-3-mini-128k-instruct

INT8-CW

1024

487

46.9

21.32196162

phi-3-mini-4k-instruct

INT8-CW

1024

486.4

46.9

21.32196162

phi-3.5-vision-instruct

INT8-CW

1032

711.4

46.9

21.32196162

qwen2.5-7b-instruct

INT4-MIXED

1024

606.9

46.9

21.32196162

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

604.9

46.9

21.32196162

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

591

47

21.27659574

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

600.8

47

21.27659574

llama-3-8b-instruct

INT4-MIXED

32

70.9

47.1

21.23142251

phi-3.5-mini-instruct

INT8-CW

1024

486.1

47.1

21.23142251

qwen2-7b-instruct

INT4-MIXED

1024

612.9

47.1

21.23142251

qwen2.5-7b-instruct

INT4-MIXED

1024

613.6

47.1

21.23142251

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

564.9

47.1

21.23142251

qwen3-4b

INT8-CW

32

59.7

47.1

21.23142251

llama-3-8b-instruct

INT4-MIXED

32

70.5

47.2

21.18644068

llama-3.1-8b-instruct

INT4-MIXED

32

70.4

47.2

21.18644068

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

641.4

47.3

21.14164905

fara-7b

INT4-MIXED

1024

762.5

47.4

21.09704641

deepseek-r1-distill-llama-8b

INT4-MIXED

35

72

47.5

21.05263158

llama-3-8b-instruct

INT4-MIXED

32

70.9

47.7

20.96436059

phi-4-mini-instruct

INT8-CW

1024

410.8

47.8

20.92050209

phi-4-mini-reasoning

INT8-CW

1024

415.2

48.1

20.79002079

afm-4.5b

INT8-CW

32

62.1

48.2

20.74688797

bloomz-7b1

INT4-MIXED

32

77.2

48.2

20.74688797

bloomz-7b1

INT4-MIXED

1024

725.3

48.3

20.70393375

qwen3-8b

INT4-MIXED

32

70.7

48.3

20.70393375

minicpm4-8b

INT4-MIXED

1024

587

48.4

20.66115702

qwen3-8b

INT4-MIXED

1024

561.2

48.4

20.66115702

falcon-7b-instruct

INT4-MIXED

1024

675.9

48.7

20.5338809

qwen-7b-chat

INT4-MIXED

1024

573.3

48.7

20.5338809

minicpm-v-4_5

INT4-MIXED

217

535.4

49

20.40816327

minicpm4-8b

INT4-MIXED

32

73.9

49

20.40816327

llama-3-8b-instruct

INT4-MIXED

1024

561.5

49.1

20.36659878

qwen3-8b

INT4-MIXED

32

72.1

49.1

20.36659878

llama-3-8b-instruct

INT4-MIXED

1024

561.5

49.2

20.32520325

llama-3.1-8b-instruct

INT4-MIXED

1024

562.5

49.3

20.28397566

minicpm4-8b

INT4-MIXED

32

75.2

49.5

20.2020202

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

629.6

49.6

20.16129032

afm-4.5b

INT8-CW

1024

385.3

49.8

20.08032129

llama-3-8b-instruct

INT4-MIXED

1024

647.9

49.8

20.08032129

qwen3-4b

INT8-CW

1024

376.9

49.9

20.04008016

llava-next-video-7b-hf

INT4-MIXED

2945

542.7

50

20

qwen3.5-9b

INT4-MIXED

83

158.6

50.3

19.88071571

qwen3-8b

INT4-MIXED

1024

597.1

50.6

19.76284585

minicpm4-8b

INT4-MIXED

1024

643.8

50.7

19.72386588

bloomz-7b1

INT4-MIXED

1024

853.9

50.8

19.68503937

qwen3.5-9b

INT4-MIXED

1024

827

51.1

19.56947162

qwen3-8b

INT4-MIXED

1024

677

51.2

19.53125

minicpm4-8b

INT4-MIXED

1024

725.5

51.4

19.45525292

zephyr-7b-beta

INT4-MIXED

32

67.8

51.7

19.34235977

baichuan2-7b-chat

INT4-MIXED

32

69.6

52.5

19.04761905

gemma-7b-it

INT4-MIXED

32

77.7

52.5

19.04761905

qwen3-vl-4b-instruct

INT8-CW

4907

1088.8

53.1

18.83239171

qwen3.5-9b

INT4-MIXED

83

164.7

53.1

18.83239171

gemma-4-e2b-it

FP16

274

343.5

53.2

18.79699248

qwen3-vl-4b-instruct

INT8-CW

4937

1123.2

53.3

18.76172608

minicpm3-4b

INT8-CW

32

133.4

53.6

18.65671642

gemma-4-e2b-it

FP16

1024

584.6

53.7

18.62197393

qwen3.5-9b

INT4-MIXED

1024

1022.5

53.8

18.58736059

zephyr-7b-beta

INT4-MIXED

1024

579.3

54

18.51851852

phi-4-multimodal-instruct

INT8-CW

578

490

54.4

18.38235294

phi-4-multimodal-instruct

INT8-CW

786

336.4

54.7

18.28153565

gemma-7b-it

INT4-MIXED

1024

645.2

54.9

18.21493625

baichuan2-7b-chat

INT4-MIXED

1024

665.1

55.3

18.08318264

gemma-7b-it

INT4-MIXED

32

88.8

55.5

18.01801802

phi-4-multimodal-instruct

INT8-CW

1362

1156.9

55.9

17.88908766

phi-4-multimodal-instruct

INT8-CW

1570

638.5

56.1

17.82531194

videochat-flash-qwen2_5-7b_internvideo2-1b

INT4-MIXED

14531

49981.7

56.4

17.73049645

gemma-7b-it

INT4-MIXED

1024

785

57.9

17.27115717

deepseek-r1-distill-llama-8b

INT4-MIXED

35

78.7

58

17.24137931

llama-3.1-8b-instruct

INT4-MIXED

32

76

58

17.24137931

gemma-2-9b-it

INT4-MIXED

32

78.6

58.8

17.00680272

minicpm3-4b

INT8-CW

1024

826.3

59.1

16.92047377

phi-2

FP16

32

65.6

59.4

16.83501684

stable-zephyr-3b-dpo

FP16

32

65.6

59.8

16.72240803

gemma-2-2b

FP16

33

64.6

60

16.66666667

stablelm-3b-4e1t

FP16

32

65.6

60

16.66666667

llama-3.1-8b-instruct

INT4-MIXED

1024

714.9

60.4

16.55629139

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

783.5

60.6

16.50165017

gemma-2-9b-it

INT4-MIXED

32

82.9

61

16.39344262

gemma-3n-e4b-it

INT8-CW

275

577.4

61.2

16.33986928

qwen-7b-chat

INT4-MIXED

32

76

61.4

16.28664495

gemma-3n-e4b-it

INT8-CW

1025

1189.3

61.5

16.2601626

gemma-2-2b

FP16

1025

355.3

61.6

16.23376623

gemma-2-9b-it

INT4-MIXED

32

83.8

61.7

16.20745543

stable-diffusion-xl-1.0-inpainting-0.1

FP16

32

62.6

62

16.12903226

gemma-2-9b-it

INT4-MIXED

1024

713.8

62.2

16.07717042

gemma-2b-it

FP16

32

66.1

62.4

16.02564103

lcm-dreamshaper-v7

INT8-HYBRID

32

64.5

62.9

15.89825119

lcm-dreamshaper-v7

INT8-HYBRID

1024

64.1

63

15.87301587

gemma-2b-it

FP16

1024

258.6

63.2

15.82278481

phi-2

FP16

1024

431.8

63.2

15.82278481

stable-zephyr-3b-dpo

FP16

1024

412

63.6

15.72327044

stablelm-3b-4e1t

FP16

1024

417.3

64

15.625

qwen-7b-chat

INT4-MIXED

1024

577.1

64.3

15.55209953

gemma-2-9b-it

INT4-MIXED

1024

766.9

64.6

15.47987616

gemma-2-9b-it

INT4-MIXED

1024

816.5

65.1

15.3609831

qwen3_8b_draft_eagle3

INT8-CW

1024

642.4

66.1

15.12859304

gemma-3n-e2b-it

FP16

275

567.1

66.2

15.10574018

gemma-3n-e2b-it

FP16

1025

1029.1

66.7

14.99250375

llama-3.2-3b-instruct

FP16

32

73.6

68.7

14.55604076

qwen3_8b_draft_eagle3

INT8-CW

32

104.3

69.4

14.4092219

llama-3.2-3b-instruct

FP16

1024

432.7

70.8

14.12429379

dolly-v2-12b

INT4-MIXED

32

114.2

71.2

14.04494382

gemma-3-12b-it

INT4-MIXED

32

105.9

73.3

13.6425648

llama-2-13b-chat-hf

INT4-MIXED

32

108.1

73.3

13.6425648

qwen2.5-coder-3b-instruct

FP16

32

78

73.9

13.53179973

smollm3-3b

FP16

32

78.5

74.2

13.47708895

dolly-v2-12b

INT4-MIXED

1024

1262.2

75.4

13.26259947

gemma-4-12b-it

INT4-MIXED

278

557.8

75.4

13.26259947

qwen2.5-coder-3b-instruct

FP16

1024

320.8

75.6

13.22751323

smollm3-3b

FP16

1024

323.3

75.7

13.21003963

qwen2.5-7b-instruct

INT8-CW

32

97.7

76.4

13.08900524

qwen2.5-7b-instruct-1m

INT8-CW

32

97.9

76.5

13.07189542

deepseek-r1-distill-qwen-7b

INT8-CW

32

98.6

76.6

13.05483029

gemma-3-12b-it

INT4-MIXED

1024

966.5

76.6

13.05483029

qwen2-7b-instruct

INT8-CW

32

98

76.6

13.05483029

qwen2.5-7b-instruct

INT8-CW

32

98.4

76.6

13.05483029

falcon-7b-instruct

INT8-CW

32

93.4

76.7

13.03780965

fara-7b

INT8-CW

32

141.2

76.7

13.03780965

lcm-dreamshaper-v7

INT8-CW

32

77.9

76.7

13.03780965

minicpm-o-2_6

INT8-CW

238

535.7

76.7

13.03780965

minicpm-v-2_6

INT8-CW

228

563.6

76.7

13.03780965

lcm-dreamshaper-v7

INT8-CW

1024

77.6

77

12.98701299

qwen2.5-vl-7b-instruct

INT8-CW

32

102.4

77.1

12.97016861

gemma-3-12b-it

INT4-MIXED

32

117.3

77.2

12.95336788

lcm-dreamshaper-v7

FP16

32

78.1

77.4

12.91989664

granite-4.0-h-micro

FP16

33

126.2

77.5

12.90322581

lcm-dreamshaper-v7

FP16

1024

78.1

77.6

12.88659794

lcm-dreamshaper-v7

INT8-CW

1024

78.3

77.6

12.88659794

lcm-dreamshaper-v7

INT8-CW

32

78.3

77.8

12.85347044

llama-2-13b-chat-hf

INT4-MIXED

1024

967.7

77.8

12.85347044

granite-4.0-h-micro

FP16

1033

879.9

77.9

12.83697047

gemma-4-12b-it

INT4-MIXED

1024

1132.6

78

12.82051282

llama-2-7b-chat-hf

INT8-CW

32

91.1

78

12.82051282

codellama-7b

INT8-CW

32

90.5

78.1

12.80409731

deepseek-r1-distill-qwen-7b

INT8-CW

1024

583.9

78.3

12.77139208

qwen2-7b-instruct

INT8-CW

1024

586.2

78.3

12.77139208

qwen2.5-7b-instruct

INT8-CW

1024

588.7

78.3

12.77139208

qwen2.5-7b-instruct-1m

INT8-CW

1024

592.5

78.3

12.77139208

qwen2.5-7b-instruct

INT8-CW

1024

597.5

78.4

12.75510204

fara-7b

INT8-CW

1024

756.6

78.6

12.72264631

falcon-7b-instruct

INT8-CW

1024

829.3

78.7

12.7064803

qwen2.5-vl-7b-instruct

INT8-CW

1024

621.1

78.7

12.7064803

phi-3-mini-4k-instruct

FP16

32

86

79.7

12.54705144

phi-3.5-mini-instruct

FP16

32

85.9

79.9

12.51564456

phi-3-mini-128k-instruct

FP16

32

86.4

80

12.5

gemma-3-12b-it

INT4-MIXED

1024

1098.8

80.2

12.46882793

llama-2-7b-chat-hf

INT8-CW

1024

578.7

81.8

12.22493888

qwen3-vl-4b-instruct_draft_eagle3

INT4-MIXED

4907

1103

82.1

12.18026797

phi-4-mini-instruct

FP16

32

90.5

82.2

12.16545012

codellama-7b

INT8-CW

1024

589.4

82.3

12.15066829

neural-chat-7b-v3-3

INT8-CW

32

97.2

82.3

12.15066829

phi-4

INT4-MIXED

32

124

82.3

12.15066829

phi-4-mini-reasoning

FP16

32

90.6

82.3

12.15066829

phi-4-reasoning

INT4-MIXED

32

123.8

82.4

12.13592233

qwen3-vl-4b-instruct_draft_eagle3

INT4-MIXED

4937

1153.2

82.7

12.09189843

baichuan2-7b-chat

INT8-CW

32

97.2

82.8

12.07729469

biomistral-7b-slerp

INT8-CW

7

86.4

83

12.04819277

phi-3.5-vision-instruct

FP16

802

298.8

83.3

12.00480192

qwen-7b-chat

INT8-CW

32

96.5

83.3

12.00480192

gemma-4-12b-it

INT4-MIXED

278

612.6

83.5

11.9760479

mistral-7b-instruct-v0.2

INT8-CW

32

96.1

83.5

11.9760479

mistral-7b-instruct-v0.3

INT8-CW

32

96

83.5

11.9760479

zephyr-7b-beta

INT8-CW

32

96.2

83.5

11.9760479

phi-3-mini-4k-instruct

FP16

1024

610.8

84.1

11.89060642

phi-3.5-mini-instruct

FP16

1024

607.8

84.1

11.89060642

mistral-7b-instruct-v0.1

INT8-CW

32

97.2

84.2

11.87648456

phi-3-mini-128k-instruct

FP16

1024

609.4

84.3

11.8623962

gemma-3-4b-it

FP16

32

93.6

84.4

11.84834123

phi-3.5-vision-instruct

FP16

1032

406.5

84.4

11.84834123

phi-4-mini-instruct

FP16

1024

581.7

84.4

11.84834123

phi-4-mini-reasoning

FP16

1024

593.9

84.5

11.83431953

qwen1.5-14b-chat

INT4-MIXED

32

122

84.5

11.83431953

neural-chat-7b-v3-3

INT8-CW

1024

620.7

85

11.76470588

bloomz-7b1

INT8-CW

32

96.8

85.3

11.72332943

phi-4

INT4-MIXED

1024

1099.2

85.6

11.68224299

phi-4-reasoning

INT4-MIXED

1024

1079.2

85.7

11.66861144

gemma-4-12b-it

INT4-MIXED

1024

1362.9

85.9

11.64144354

mistral-7b-instruct-v0.2

INT8-CW

1024

608.4

85.9

11.64144354

mistral-7b-instruct-v0.3

INT8-CW

1024

611.2

85.9

11.64144354

zephyr-7b-beta

INT8-CW

1024

605.9

86

11.62790698

gemma-3-4b-it

FP16

1024

509.9

86.2

11.60092807

phi-4-reasoning

INT4-MIXED

32

129.3

86.2

11.60092807

baichuan2-7b-chat

INT8-CW

1024

647.1

86.6

11.54734411

llama-3.1-8b-instruct

INT8-CW

32

100.8

86.7

11.53402537

mistral-7b-instruct-v0.1

INT8-CW

1025

648.7

86.9

11.50747986

deepseek-r1-distill-llama-8b

INT8-CW

35

114.1

87

11.49425287

phi-4

INT4-MIXED

32

130.5

87.1

11.48105626

qwen-7b-chat

INT8-CW

1024

613.5

87.1

11.48105626

llama-3-8b-instruct

INT8-CW

32

101.4

87.2

11.46788991

deepseek-r1-distill-qwen-14b

INT4-MIXED

32

131.5

88.3

11.32502831

qwen3-4b

FP16

32

94.9

88.5

11.29943503

bloomz-7b1

INT8-CW

1024

826.1

88.8

11.26126126

qwen1.5-14b-chat

INT4-MIXED

1024

1140.1

89

11.23595506

llama-3.1-8b-instruct

INT8-CW

1024

614.2

89.1

11.22334456

videochat-flash-qwen2_5-7b_internvideo2-1b

INT8-CW

14531

50454.6

89.2

11.21076233

phi-4-reasoning

INT4-MIXED

1024

1149

89.6

11.16071429

deepseek-r1-distill-llama-8b

INT8-CW

1135

679.6

89.7

11.14827202

llama-3-8b-instruct

INT8-CW

1024

617

89.7

11.14827202

llava-next-video-7b-hf

INT8-CW

2945

555.3

89.8

11.13585746

llava-v1.6-mistral-7b-hf

INT8-CW

2944

771.1

90.2

11.0864745

phi-4

INT4-MIXED

1024

1302.3

90.4

11.0619469

phi-4-reasoning

INT4-MIXED

32

138.6

91.2

10.96491228

qwen3-4b

FP16

1024

509.7

91.4

10.94091904

qwen3-8b

INT8-CW

32

103.3

91.4

10.94091904

llama-2-13b-chat-hf

INT4-MIXED

32

137.2

91.8

10.89324619

afm-4.5b

FP16

32

100.7

92

10.86956522

deepseek-r1-distill-qwen-14b

INT4-MIXED

1024

1205.9

92.1

10.8577633

minicpm-v-4_5

INT8-CW

217

512.8

93.3

10.71811361

starcoder

INT4-MIXED

32

152.1

93.9

10.64962726

qwen3-8b

INT8-CW

1024

652.9

94.4

10.59322034

phi-4-reasoning

INT4-MIXED

1024

1369.5

94.5

10.58201058

afm-4.5b

FP16

1024

540.1

94.9

10.5374078

minicpm3-4b

FP16

32

112.6

95

10.52631579

phi-4-multimodal-instruct

FP16

578

647.8

95.3

10.49317943

phi-4-multimodal-instruct

FP16

786

454.6

95.5

10.47120419

gemma-7b-it

INT8-CW

32

117.9

95.8

10.43841336

qwen3.5-9b

INT8-CW

83

188

96.1

10.40582726

minicpm4-8b

INT8-CW

32

108.8

96.3

10.38421599

llama-2-13b-chat-hf

INT4-MIXED

1024

1158.3

96.6

10.35196687

phi-4-multimodal-instruct

FP16

1362

1504.3

96.8

10.33057851

starcoder

INT4-MIXED

1024

1436.3

97

10.30927835

qwen3.5-9b

INT8-CW

1024

849.2

97.1

10.29866117

phi-4-multimodal-instruct

FP16

1570

786.5

97.2

10.28806584

minicpm4-8b

INT8-CW

1024

700.2

98.5

10.15228426

gemma-7b-it

INT8-CW

1024

778.8

99.9

10.01001001

qwen3-vl-4b-instruct

FP16

4907

1328.9

100

10

Topology

Precision

Input Size

1st latency (ms)

2nd latency (ms)

2nd token per sec

minicpm4-0.5b

INT4-MIXED

32

21.1

7.3

136.9863014

minicpm4-0.5b

INT4-MIXED

32

20.9

7.5

133.3333333

gemma-3-270m

INT4-MIXED

1024

64.9

7.6

131.5789474

gemma-3-270m

INT4-MIXED

32

21

7.6

131.5789474

minicpm4-0.5b

INT4-MIXED

32

21.9

7.6

131.5789474

lfm2.5-350m

INT4-MIXED

32

22.8

7.9

126.5822785

gemma-3-270m

INT8-CW

1024

60.5

8

125

gemma-3-270m

INT8-CW

32

21.3

8

125

lfm2.5-350m

INT4-MIXED

1024

153.9

8.2

121.9512195

minicpm4-0.5b

INT4-MIXED

1024

98.9

8.4

119.047619

minicpm4-0.5b

INT4-MIXED

1024

101.8

8.5

117.6470588

minicpm4-0.5b

INT4-MIXED

1024

127.3

8.7

114.9425287

nuextract-tiny

INT4-MIXED

32

21.8

8.9

112.3595506

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

21

8.9

112.3595506

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

21.9

9

111.1111111

qwen2.5-coder-0.5b-instruct

INT4-MIXED

32

21.9

9.2

108.6956522

lfm2.5-350m

INT8-CW

32

26

9.7

103.0927835

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

104.3

9.8

102.0408163

nuextract-tiny

INT4-MIXED

1024

122.2

9.9

101.010101

lfm2.5-350m

INT8-CW

1024

140.1

10

100

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

106.9

10

100

qwen2.5-coder-0.5b-instruct

INT4-MIXED

1024

134.2

10.2

98.03921569

minicpm4-0.5b

INT8-CW

32

23

10.4

96.15384615

whisper-small

INT4-MIXED

prompt0

133.3

10.4

96.15384615

nanollava

INT4-MIXED

760

151.9

10.5

95.23809524

whisper-small

INT4-MIXED

prompt1

199.1

10.6

94.33962264

whisper-small

INT4-MIXED

prompt1

186.8

10.8

92.59259259

whisper-small

INT4-MIXED

prompt0

134.7

11

90.90909091

nuextract-tiny

INT8-CW

32

22.9

11.1

90.09009009

whisper-small

INT4-MIXED

prompt1

201.6

11.3

88.49557522

whisper-small

INT8-CW

prompt0

141.4

11.3

88.49557522

distil-large-v2

INT8-CW

prompt0

387.7

11.4

87.71929825

distil-large-v2

INT8-CW

prompt1

440

11.4

87.71929825

whisper-small

INT4-MIXED

prompt0

149.1

11.4

87.71929825

whisper-small

INT8-CW

prompt1

197.9

11.4

87.71929825

distil-large-v2

INT4-MIXED

prompt0

430

11.5

86.95652174

minicpm4-0.5b

INT8-CW

1024

109.9

11.5

86.95652174

distil-large-v2

INT4-MIXED

prompt1

491.6

11.7

85.47008547

nanollava

INT4-MIXED

1752

317.7

11.7

85.47008547

qwen2.5-coder-0.5b-instruct

INT8-CW

32

23.5

11.9

84.03361345

whisper-large-v3-turbo

INT8-CW

prompt1

464.5

12.1

82.6446281

gemma-3-270m

FP16

32

21.4

12.2

81.96721311

gemma-3-270m

FP16

1024

69.7

12.3

81.30081301

nanollava

INT8-CW

760

154.2

12.3

81.30081301

whisper-large-v3-turbo

INT8-CW

prompt0

403.4

12.4

80.64516129

nuextract-tiny

INT8-CW

1024

104.3

12.5

80

whisper-large-v3-turbo

INT4-MIXED

prompt1

511.1

12.6

79.36507937

qwen2.5-coder-0.5b-instruct

INT8-CW

1024

116.5

12.7

78.74015748

whisper-large-v3-turbo

INT4-MIXED

prompt0

441.9

12.7

78.74015748

tiny-llama-1.1b-chat

INT4-MIXED

32

29.2

13

76.92307692

nanollava

INT8-CW

1752

292.7

13.9

71.94244604

qwen3.5-0.8b

INT4-MIXED

85

57.1

13.9

71.94244604

minicpm5-1b

INT4-MIXED

32

27.3

14

71.42857143

qwen3.5-0.8b

INT4-MIXED

85

57

14.1

70.92198582

qwen3.5-0.8b

INT4-MIXED

85

63.2

14.3

69.93006993

qwen3-asr-0.6b

INT4-MIXED

prompt0

225.2

14.4

69.44444444

qwen3-asr-0.6b

INT4-MIXED

prompt0

224.7

14.4

69.44444444

qwen3.5-0.8b

INT4-MIXED

1024

337

14.4

69.44444444

whisper-small

FP16

prompt1

197.8

14.5

68.96551724

qwen3-asr-0.6b

INT4-MIXED

prompt0

235.7

14.6

68.49315068

whisper-small

FP16

prompt0

138

14.6

68.49315068

qwen3.5-0.8b

INT4-MIXED

1024

338.1

14.8

67.56756757

tiny-llama-1.1b-chat

INT4-MIXED

1024

217

14.8

67.56756757

qwen3.5-0.8b

INT4-MIXED

1024

390.3

15

66.66666667

qwen3-asr-0.6b

INT4-MIXED

prompt1

680.7

15.3

65.35947712

lfm2.5-1.2b-instruct

INT4-MIXED

32

36.1

15.4

64.93506494

qwen3-asr-0.6b

INT4-MIXED

prompt1

613.6

15.4

64.93506494

lfm2-1.2b

INT4-MIXED

32

35.4

15.6

64.1025641

tiny-llama-1.1b-chat

INT4-MIXED

32

33

15.7

63.69426752

distil-large-v2

FP16

prompt0

433.3

15.8

63.29113924

distil-large-v2

FP16

prompt1

487.8

15.8

63.29113924

lfm2-1.2b

INT4-MIXED

1024

319.7

15.9

62.89308176

lfm2-1.2b

INT4-MIXED

32

40.6

15.9

62.89308176

lfm2.5-1.2b-instruct

INT4-MIXED

1024

362

15.9

62.89308176

lfm2.5-1.2b-instruct

INT4-MIXED

32

40.5

15.9

62.89308176

qwen3-asr-0.6b

INT8-CW

prompt0

231.8

15.9

62.89308176

minicpm5-1b

INT4-MIXED

1024

222.8

16

62.5

qwen3-asr-0.6b

INT4-MIXED

prompt1

603.5

16

62.5

lfm2.5-1.2b-instruct

INT4-MIXED

32

40.6

16.1

62.11180124

lfm2.5-350m

FP16

32

26.3

16.1

62.11180124

lfm2-1.2b

INT4-MIXED

32

40.8

16.2

61.72839506

lfm2-1.2b

INT4-MIXED

1024

361.7

16.3

61.34969325

llama-3.2-1b-instruct

INT4-MIXED

32

32.4

16.3

61.34969325

whisper-large-v3-turbo

FP16

prompt0

449

16.3

61.34969325

whisper-large-v3-turbo

FP16

prompt1

514.2

16.3

61.34969325

lfm2.5-1.2b-instruct

INT4-MIXED

1024

324.9

16.4

60.97560976

lfm2.5-350m

FP16

1024

159.9

16.4

60.97560976

llama-3.2-1b-instruct

INT4-MIXED

32

32.9

16.5

60.60606061

lfm2-1.2b

INT4-MIXED

1024

424.6

16.6

60.24096386

lfm2.5-1.2b-instruct

INT4-MIXED

1024

399

16.6

60.24096386

gemma-3-1b-it

INT4-MIXED

32

30.6

17

58.82352941

exaone-4.0-1.2b

INT4-MIXED

32

32.9

17.3

57.80346821

gemma-3-1b-it

INT4-MIXED

1024

200.3

17.4

57.47126437

tiny-llama-1.1b-chat

INT4-MIXED

1024

295.1

17.5

57.14285714

llama-3.2-1b-instruct

INT4-MIXED

1024

209.5

17.7

56.49717514

exaone-4.0-1.2b

INT4-MIXED

32

37.1

17.8

56.17977528

gemma-3-1b-it

INT4-MIXED

32

33

17.8

56.17977528

exaone-4.0-1.2b

INT4-MIXED

32

37.7

18

55.55555556

qwen3.5-0.8b

INT8-CW

85

64.8

18

55.55555556

gemma-3-1b-it

INT4-MIXED

32

33.6

18.1

55.24861878

llama-3.2-1b-instruct

INT4-MIXED

1024

254.3

18.1

55.24861878

qwen3-asr-0.6b

INT8-CW

prompt1

632.6

18.2

54.94505495

gemma-3-1b-it

INT4-MIXED

1024

205

18.3

54.64480874

gemma-3-1b-it

INT4-MIXED

1024

242.4

18.5

54.05405405

qwen3.5-0.8b

INT8-CW

1024

347.7

18.8

53.19148936

minicpm4-0.5b

FP16

32

23.7

19.3

51.8134715

minicpm5-1b

INT8-CW

32

34.9

19.3

51.8134715

nuextract-tiny

FP16

32

25

19.3

51.8134715

minicpm-1b-sft

INT4-MIXED

31

43.7

19.6

51.02040816

exaone-4.0-1.2b

INT4-MIXED

1024

330.3

19.8

50.50505051

exaone-4.0-1.2b

INT4-MIXED

1024

341

20.3

49.26108374

minicpm4-0.5b

FP16

1024

133.3

20.3

49.26108374

exaone-4.0-1.2b

INT4-MIXED

1024

371.8

20.5

48.7804878

minicpm-1b-sft

INT4-MIXED

31

45.3

20.5

48.7804878

minicpm5-1b

INT8-CW

1024

201.4

20.7

48.30917874

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

39.5

20.8

48.07692308

minicpm-1b-sft

INT4-MIXED

31

46.2

20.9

47.84688995

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

42.2

21.1

47.39336493

nanollava

FP16

760

186

21.3

46.94835681

nuextract-tiny

FP16

1024

128.8

21.4

46.72897196

tiny-llama-1.1b-chat

INT8-CW

32

38.4

21.4

46.72897196

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

44.5

21.5

46.51162791

qwen2.5-1.5b-instruct

INT4-MIXED

32

43.9

21.6

46.2962963

qwen2.5-coder-0.5b-instruct

FP16

32

26.7

21.8

45.87155963

qwen2.5-coder-1.5b-instruct

INT4-MIXED

32

44.8

21.9

45.66210046

qwen2.5-coder-0.5b-instruct

FP16

1024

143.1

22.4

44.64285714

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

283.4

22.6

44.24778761

minicpm-1b-sft

INT4-MIXED

1014

316.1

22.7

44.05286344

gemma-3-1b-it

INT8-CW

32

39.2

22.8

43.85964912

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

343.3

22.9

43.66812227

gemma-3-1b-it

INT8-CW

1024

230.1

23.3

42.91845494

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

342.1

23.3

42.91845494

tiny-llama-1.1b-chat

INT8-CW

1024

237.9

23.3

42.91845494

nanollava

FP16

1752

329.8

23.4

42.73504274

qwen2.5-1.5b-instruct

INT4-MIXED

32

47.3

23.4

42.73504274

qwen2.5-1.5b-instruct

INT4-MIXED

1024

343.5

23.4

42.73504274

minicpm-1b-sft

INT4-MIXED

1014

326.8

23.7

42.19409283

qwen2.5-coder-1.5b-instruct

INT4-MIXED

1024

358.6

23.8

42.01680672

minicpm-1b-sft

INT4-MIXED

1014

380.9

24.1

41.49377593

lfm2-8b-a1b

INT4-MIXED

32

382.4

24.6

40.6504065

lfm2.5-1.2b-instruct

INT4-MIXED

32

48.1

25

40

lfm2-1.2b

INT4-MIXED

32

48.9

25.1

39.84063745

lfm2-8b-a1b

INT4-MIXED

1024

895.3

25.1

39.84063745

llama-3.2-1b-instruct

INT8-CW

32

43.3

25.1

39.84063745

qwen2.5-1.5b-instruct

INT4-MIXED

1024

367

25.3

39.5256917

lfm2-1.2b

INT8-CW

32

50.9

25.5

39.21568627

lfm2.5-1.2b-instruct

INT8-CW

32

53.2

25.5

39.21568627

lfm2.5-1.2b-instruct

INT4-MIXED

1024

344

25.6

39.0625

lfm2-1.2b

INT4-MIXED

1024

366.2

25.7

38.91050584

lfm2-1.2b

INT8-CW

1024

361.5

26

38.46153846

lfm2.5-1.2b-instruct

INT8-CW

1024

368.4

26

38.46153846

qwen3-asr-0.6b

FP16

prompt0

269.1

26.3

38.02281369

llama-3.2-1b-instruct

INT8-CW

1024

237.2

26.7

37.45318352

exaone-4.0-1.2b

INT8-CW

32

47.4

27.1

36.900369

qwen3-asr-0.6b

FP16

prompt1

709.2

29.1

34.36426117

exaone-4.0-1.2b

INT8-CW

1024

356.5

29.7

33.67003367

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

32

59.5

30.1

33.22259136

qwen3.5-0.8b

FP16

85

71.3

31.2

32.05128205

minicpm-1b-sft

INT8-CW

31

59.6

31.5

31.74603175

qwen3.5-0.8b

FP16

1024

387.3

31.7

31.54574132

deepseek-r1-distill-qwen-1.5b

INT4-MIXED

1024

347.4

31.8

31.44654088

deepseek-r1-distill-qwen-1.5b

INT8-CW

32

58.9

32.5

30.76923077

qwen2.5-1.5b-instruct

INT8-CW

32

61.5

33

30.3030303

qwen2.5-coder-1.5b-instruct

INT8-CW

32

61.4

33.1

30.21148036

phi-2

INT4-MIXED

32

66

33.3

30.03003003

stable-zephyr-3b-dpo

INT4-MIXED

32

62.4

33.7

29.6735905

deepseek-r1-distill-qwen-1.5b

INT8-CW

1024

318.9

33.8

29.58579882

stablelm-3b-4e1t

INT4-MIXED

32

63.1

34

29.41176471

gemma-2b-it

INT4-MIXED

32

55.6

34.1

29.3255132

gemma-4-e2b-it

INT4-MIXED

274

601.8

34.1

29.3255132

qwen2.5-coder-1.5b-instruct

INT8-CW

1024

332.8

34.4

29.06976744

minicpm-1b-sft

INT8-CW

1014

357.1

34.5

28.98550725

qwen2.5-1.5b-instruct

INT8-CW

1024

333.7

34.5

28.98550725

gemma-2b-it

INT4-MIXED

32

65.4

35.1

28.49002849

gemma-4-e2b-it

INT4-MIXED

1024

962.2

35.3

28.3286119

gemma-2b-it

INT4-MIXED

1024

381.6

35.6

28.08988764

minicpm5-1b

FP16

32

43

36.1

27.70083102

whisper-large-v3

INT4-MIXED

prompt0

566.6

36.2

27.62430939

gemma-2-2b

INT4-MIXED

33

57.1

36.4

27.47252747

gemma-2b-it

INT4-MIXED

1024

477.5

36.6

27.32240437

minicpm5-1b

FP16

1024

256.5

37.6

26.59574468

stable-zephyr-3b-dpo

INT4-MIXED

32

78.2

37.7

26.52519894

whisper-large-v3

INT4-MIXED

prompt1

620.4

37.8

26.45502646

phi-2

INT4-MIXED

1024

701.8

37.9

26.38522427

gemma-4-e2b-it

INT4-MIXED

274

651.3

38.3

26.10966057

stable-zephyr-3b-dpo

INT4-MIXED

1024

785.2

38.3

26.10966057

stablelm-3b-4e1t

INT4-MIXED

1024

788.4

38.3

26.10966057

gemma-2-2b

INT4-MIXED

33

58.4

38.7

25.83979328

gemma-2-2b

INT4-MIXED

1025

514.4

38.7

25.83979328

gemma-4-e2b-it

INT4-MIXED

1024

1107.4

39.4

25.38071066

smollm3-3b

INT4-MIXED

32

69.6

39.4

25.38071066

gemma-3n-e2b-it

INT4-MIXED

275

1043.8

39.5

25.3164557

phi-2

INT4-MIXED

32

84.6

39.6

25.25252525

llama-3.2-3b-instruct

INT4-MIXED

32

69.3

39.9

25.06265664

gemma-3n-e2b-it

INT4-MIXED

1025

2136.4

40

25

qwen2.5-coder-3b-instruct

INT4-MIXED

32

69.9

40.2

24.87562189

smollm3-3b

INT4-MIXED

32

80.8

40.4

24.75247525

tiny-llama-1.1b-chat

FP16

32

49.1

40.7

24.57002457

gemma-3n-e2b-it

INT4-MIXED

275

1034.1

40.8

24.50980392

gemma-2-2b

INT4-MIXED

1025

587.9

41

24.3902439

granite-4.0-h-micro

INT4-MIXED

33

109.1

41

24.3902439

smollm3-3b

INT4-MIXED

32

83.3

41

24.3902439

llama-3.2-3b-instruct

INT4-MIXED

32

80.6

41.1

24.33090024

qwen2.5-coder-3b-instruct

INT4-MIXED

32

81.2

41.1

24.33090024

granite-4.0-h-micro

INT4-MIXED

1033

1903.9

41.4

24.15458937

llama-3.2-3b-instruct

INT4-MIXED

32

81.3

41.4

24.15458937

gemma-3n-e2b-it

INT4-MIXED

1025

2186.8

41.6

24.03846154

gemma-3-1b-it

FP16

32

51.2

42.2

23.69668246

tiny-llama-1.1b-chat

FP16

1024

319.5

42.2

23.69668246

whisper-large-v3

INT8-CW

prompt0

514

42.2

23.69668246

stable-zephyr-3b-dpo

INT4-MIXED

1024

871

42.3

23.64066194

granite-4.0-h-micro

INT4-MIXED

33

110.1

42.4

23.58490566

gemma-3-1b-it

FP16

1024

292.3

42.6

23.4741784

smollm3-3b

INT4-MIXED

1024

529.5

42.6

23.4741784

stablelm-3b-4e1t

INT4-MIXED

32

88.4

42.6

23.4741784

granite-4.0-h-micro

INT4-MIXED

33

110.6

42.7

23.41920375

granite-4.0-h-micro

INT4-MIXED

1033

1971.2

42.9

23.31002331

whisper-large-v3

INT8-CW

prompt1

571.1

42.9

23.31002331

granite-4.0-h-micro

INT4-MIXED

1033

2047.2

43.1

23.20185615

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

504.2

43.4

23.04147465

phi-3-mini-128k-instruct

INT4-MIXED

32

81

43.5

22.98850575

phi-3.5-mini-instruct

INT4-MIXED

32

80

43.5

22.98850575

phi-3-mini-4k-instruct

INT4-MIXED

32

80.4

43.6

22.93577982

smollm3-3b

INT4-MIXED

1024

566.2

43.6

22.93577982

llama-3.2-3b-instruct

INT4-MIXED

1024

550.4

43.7

22.88329519

smollm3-3b

INT4-MIXED

1024

671.5

44

22.72727273

phi-2

INT4-MIXED

1024

951.2

44.3

22.57336343

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

528

44.3

22.57336343

llama-3.2-3b-instruct

INT4-MIXED

1024

592.9

44.7

22.37136465

phi-3-mini-4k-instruct

INT4-MIXED

32

96.8

44.8

22.32142857

llama-3.2-3b-instruct

INT4-MIXED

1024

677.2

44.9

22.27171492

phi-3.5-mini-instruct

INT4-MIXED

32

95.8

45

22.22222222

phi-3-mini-128k-instruct

INT4-MIXED

32

100.8

45.6

21.92982456

qwen2.5-3b-instruct-gptq

INT4-MIXED

32

87.4

46.7

21.41327623

stablelm-3b-4e1t

INT4-MIXED

1024

893

47.2

21.18644068

lfm2-1.2b

FP16

32

63.8

47.8

20.92050209

qwen2.5-coder-3b-instruct

INT4-MIXED

32

95.4

48

20.83333333

lfm2.5-1.2b-instruct

FP16

32

64.4

48.1

20.79002079

lfm2-1.2b

FP16

1024

471.9

48.3

20.70393375

phi-3-mini-128k-instruct

INT4-MIXED

1024

863.5

48.4

20.66115702

phi-3.5-mini-instruct

INT4-MIXED

1024

867.6

48.4

20.66115702

llama-3.2-1b-instruct

FP16

32

57.5

48.5

20.6185567

phi-3-mini-4k-instruct

INT4-MIXED

1024

861.1

48.5

20.6185567

lfm2.5-1.2b-instruct

FP16

1024

534.2

48.6

20.57613169

qwen2.5-3b-instruct-gptq

INT4-MIXED

1024

552.5

48.9

20.44989775

gemma-3-4b-it

INT4-MIXED

32

85.9

49.6

20.16129032

phi-3-mini-4k-instruct

INT4-MIXED

1024

907.6

49.6

20.16129032

phi-3.5-vision-instruct

INT4-MIXED

802

568.9

49.6

20.16129032

phi-4-mini-instruct

INT4-MIXED

32

88.4

49.8

20.08032129

phi-3.5-mini-instruct

INT4-MIXED

1024

910.6

49.9

20.04008016

phi-4-mini-reasoning

INT4-MIXED

32

85.8

49.9

20.04008016

llama-3.2-1b-instruct

FP16

1024

314.7

50

20

gemma-2b-it

INT8-CW

32

86.3

50.2

19.92031873

phi-3-mini-128k-instruct

INT4-MIXED

1024

1117.1

50.5

19.8019802

qwen3-4b

INT4-MIXED

32

88.9

50.6

19.76284585

phi-3.5-vision-instruct

INT4-MIXED

1032

1729.8

50.7

19.72386588

phi-4-mini-instruct

INT4-MIXED

32

100.5

51

19.60784314

gemma-3-4b-it

INT4-MIXED

32

99.5

51.1

19.56947162

phi-4-mini-reasoning

INT4-MIXED

32

99.3

51.1

19.56947162

qwen2.5-coder-3b-instruct

INT4-MIXED

1024

838.4

51.2

19.53125

gemma-4-e2b-it

INT8-CW

274

681.9

51.3

19.49317739

gemma-3-4b-it

INT4-MIXED

32

101.9

51.5

19.41747573

phi-4-mini-reasoning

INT4-MIXED

32

103.1

51.5

19.41747573

gemma-2b-it

INT8-CW

1024

454.7

51.8

19.30501931

gemma-3-4b-it

INT4-MIXED

1024

747.3

52

19.23076923

qwen3-4b

INT4-MIXED

32

103.9

52.3

19.12045889

gemma-4-e2b-it

INT8-CW

1024

1105.2

52.5

19.04761905

afm-4.5b

INT4-MIXED

32

104.1

52.6

19.01140684

phi-3-mini-4k-instruct

INT4-MIXED

32

114.5

53.1

18.83239171

phi-3.5-mini-instruct

INT4-MIXED

32

117.7

53.2

18.79699248

gemma-3-4b-it

INT4-MIXED

1024

779.9

53.5

18.69158879

gemma-3-4b-it

INT4-MIXED

1024

844.6

53.9

18.5528757

phi-4-mini-instruct

INT4-MIXED

1024

866.8

54

18.51851852

phi-4-mini-reasoning

INT4-MIXED

1024

865.9

54.1

18.48428835

exaone-4.0-1.2b

FP16

32

63.6

55.1

18.14882033

phi-4-mini-reasoning

INT4-MIXED

1024

914.8

55.2

18.11594203

phi-4-mini-instruct

INT4-MIXED

1024

909.2

55.3

18.08318264

phi-4-mini-reasoning

INT4-MIXED

1024

979.9

55.6

17.98561151

gemma-2-2b

INT8-CW

33

104.5

55.7

17.95332136

phi-2

INT8-CW

32

106.5

55.7

17.95332136

minicpm3-4b

INT4-MIXED

32

103.7

56

17.85714286

afm-4.5b

INT4-MIXED

1024

886.7

56.3

17.76198934

stable-zephyr-3b-dpo

INT8-CW

32

104.8

56.3

17.76198934

qwen3-4b

INT4-MIXED

1024

793.2

56.7

17.6366843

minicpm-1b-sft

FP16

31

68.7

56.8

17.6056338

exaone-4.0-1.2b

FP16

1024

446.7

57.8

17.30103806

phi-3-mini-4k-instruct

INT4-MIXED

1024

1240.5

57.9

17.27115717

minicpm3-4b

INT4-MIXED

32

122.3

58

17.24137931

phi-3.5-mini-instruct

INT4-MIXED

1024

1352.4

58.1

17.21170396

phi-4-mini-instruct

INT4-MIXED

32

117

58.2

17.18213058

qwen3-4b

INT4-MIXED

1024

806.3

58.3

17.15265866

gemma-2-2b

INT8-CW

1025

644

58.6

17.06484642

minicpm3-4b

INT4-MIXED

32

123

58.6

17.06484642

stablelm-3b-4e1t

INT8-CW

32

105.3

59

16.94915254

qwen3-vl-4b-instruct

INT4-MIXED

4907

2389.1

59.5

16.80672269

qwen3-vl-4b-instruct

INT4-MIXED

4937

2470.6

59.5

16.80672269

gemma-3n-e2b-it

INT8-CW

275

1103.4

60

16.66666667

minicpm-1b-sft

FP16

1014

458.6

60.2

16.61129568

phi-2

INT8-CW

1024

806.7

60.3

16.58374793

gemma-3n-e2b-it

INT8-CW

1025

2213.8

60.8

16.44736842

stable-zephyr-3b-dpo

INT8-CW

1024

886.4

61

16.39344262

qwen3-vl-4b-instruct

INT4-MIXED

4907

2380.8

61.1

16.36661211

qwen3-vl-4b-instruct

INT4-MIXED

4937

2462.7

61.2

16.33986928

deepseek-r1-distill-qwen-1.5b

FP16

32

74.1

61.3

16.3132137

qwen2.5-coder-3b-instruct

INT8-CW

32

111.6

61.6

16.23376623

qwen3-vl-4b-instruct

INT4-MIXED

4907

2376

61.6

16.23376623

qwen3-vl-4b-instruct

INT4-MIXED

4937

2464.4

61.6

16.23376623

qwen2.5-coder-1.5b-instruct

FP16

32

73.6

61.7

16.20745543

smollm3-3b

INT8-CW

32

113.6

61.7

16.20745543

qwen2.5-1.5b-instruct

FP16

32

73.9

61.8

16.18122977

phi-4-mini-instruct

INT4-MIXED

1024

1205.9

62.3

16.05136437

deepseek-r1-distill-qwen-1.5b

FP16

1024

438.4

62.7

15.94896332

phi-4-multimodal-instruct

INT4-MIXED

578

983.4

63.1

15.84786054

qwen2.5-1.5b-instruct

FP16

1024

438.3

63.2

15.82278481

qwen2.5-coder-1.5b-instruct

FP16

1024

440.5

63.2

15.82278481

stablelm-3b-4e1t

INT8-CW

1024

903.2

63.8

15.67398119

qwen2.5-coder-3b-instruct

INT8-CW

1024

620.7

64

15.625

smollm3-3b

INT8-CW

1024

634.2

64

15.625

whisper-large-v3

FP16

prompt0

581.1

64

15.625

phi-4-multimodal-instruct

INT4-MIXED

786

691.5

64.1

15.60062402

whisper-large-v3

FP16

prompt1

641.4

64.3

15.55209953

minicpm3-4b

INT4-MIXED

1024

1886.8

64.7

15.45595054

llama-3.2-3b-instruct

INT8-CW

32

114.5

65.6

15.24390244

phi-4-multimodal-instruct

INT4-MIXED

1362

2352.6

66.4

15.06024096

granite-4.0-h-micro

INT8-CW

33

168.6

66.5

15.03759398

minicpm3-4b

INT4-MIXED

1024

1935.9

66.6

15.01501502

granite-4.0-h-micro

INT8-CW

1033

2064.9

67

14.92537313

phi-4-multimodal-instruct

INT4-MIXED

1570

1352.1

67.1

14.90312966

minicpm3-4b

INT4-MIXED

1024

2070

67.3

14.85884101

llama-3.2-3b-instruct

INT8-CW

1024

683.8

68.4

14.61988304

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

74.1

71.5

13.98601399

codellama-7b

INT4-MIXED

32

123.2

71.9

13.90820584

stable-diffusion-xl-1.0-inpainting-0.1

INT8-CW

32

74.7

72

13.88888889

gemma-3n-e4b-it

INT4-MIXED

275

1100.4

72.6

13.77410468

llama-2-7b-chat-hf

INT4-MIXED

32

126.7

72.7

13.75515818

qwen3_8b_draft_eagle3

INT4-MIXED

32

192.4

73.1

13.67989056

gemma-3n-e4b-it

INT4-MIXED

1025

2665

73.3

13.6425648

biomistral-7b-slerp

INT4-MIXED

7

82.5

73.7

13.56852103

codellama-7b

INT4-MIXED

32

158.9

75.1

13.31557923

falcon-7b-instruct

INT4-MIXED

32

135.6

75.3

13.28021248

llama-2-7b-chat-hf

INT4-MIXED

32

158.8

75.3

13.28021248

gemma-3n-e4b-it

INT4-MIXED

275

1082.4

75.4

13.26259947

neural-chat-7b-v3-3

INT4-MIXED

32

135.7

75.9

13.17523057

gemma-3n-e4b-it

INT4-MIXED

1025

2794.1

76.1

13.14060447

zephyr-7b-beta

INT4-MIXED

32

135.7

76.1

13.14060447

mistral-7b-instruct-v0.2

INT4-MIXED

32

137.2

76.3

13.1061599

phi-3-mini-128k-instruct

INT8-CW

32

143.2

76.3

13.1061599

phi-3.5-mini-instruct

INT8-CW

32

142.7

76.3

13.1061599

mistral-7b-instruct-v0.3

INT4-MIXED

32

135.6

76.7

13.03780965

phi-3-mini-4k-instruct

INT8-CW

32

143.9

76.7

13.03780965

phi-4-multimodal-instruct

INT4-MIXED

578

1168.5

76.7

13.03780965

biomistral-7b-slerp

INT4-MIXED

7

86.3

77.4

12.91989664

phi-4-multimodal-instruct

INT4-MIXED

786

782.1

77.6

12.88659794

codellama-7b

INT4-MIXED

1024

1144.8

77.8

12.85347044

phi-4-mini-instruct

INT8-CW

32

137.9

78.2

12.78772379

phi-4-mini-reasoning

INT8-CW

32

137.8

78.2

12.78772379

llama-2-7b-chat-hf

INT4-MIXED

1024

1128.3

78.7

12.7064803

gemma-3-4b-it

INT8-CW

32

140.6

78.9

12.67427123

mistral-7b-instruct-v0.2

INT4-MIXED

32

169.9

79

12.65822785

mistral-7b-instruct-v0.3

INT4-MIXED

32

169.8

79

12.65822785

neural-chat-7b-v3-3

INT4-MIXED

32

171.3

79.5

12.57861635

mistral-7b-instruct-v0.1

INT4-MIXED

32

173.1

79.8

12.53132832

mistral-7b-instruct-v0.2

INT4-MIXED

32

171.6

79.8

12.53132832

mistral-7b-instruct-v0.3

INT4-MIXED

32

172.2

79.9

12.51564456

phi-4-multimodal-instruct

INT4-MIXED

1362

2769.4

80

12.5

falcon-7b-instruct

INT4-MIXED

1024

1173.1

80.5

12.42236025

qwen2.5-7b-instruct

INT4-MIXED

32

153.7

80.7

12.39157373

qwen2.5-7b-instruct-1m

INT4-MIXED

32

151.5

80.8

12.37623762

phi-4-multimodal-instruct

INT4-MIXED

1570

1453.9

80.9

12.36093943

codellama-7b

INT4-MIXED

1024

1367.4

81.1

12.33045623

llama-2-7b-chat-hf

INT4-MIXED

1024

1230.6

81.3

12.300123

neural-chat-7b-v3-3

INT4-MIXED

1024

1223.2

81.4

12.28501229

phi-4-mini-reasoning

INT8-CW

1024

1014.2

81.5

12.26993865

qwen3-4b

INT8-CW

32

146.9

81.5

12.26993865

qwen2.5-vl-7b-instruct

INT4-MIXED

32

129.5

81.6

12.25490196

deepseek-r1-distill-llama-8b

INT4-MIXED

35

128.4

81.7

12.23990208

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

170.5

81.7

12.23990208

minicpm-v-2_6

INT4-MIXED

228

934.8

81.7

12.23990208

phi-3-mini-128k-instruct

INT8-CW

1024

1155.4

81.7

12.23990208

phi-3.5-mini-instruct

INT8-CW

1024

1149

81.7

12.23990208

phi-3.5-vision-instruct

INT8-CW

802

626.8

81.7

12.23990208

phi-4-mini-instruct

INT8-CW

1024

1012

81.7

12.23990208

zephyr-7b-beta

INT4-MIXED

1024

1221

81.7

12.23990208

minicpm-o-2_6

INT4-MIXED

238

947.9

81.8

12.22493888

phi-3-mini-4k-instruct

INT8-CW

1024

1159.9

81.8

12.22493888

mistral-7b-instruct-v0.2

INT4-MIXED

1024

1220.1

81.9

12.21001221

mistral-7b-instruct-v0.3

INT4-MIXED

1024

1206.3

81.9

12.21001221

gemma-3-4b-it

INT8-CW

1024

881.9

82.1

12.18026797

qwen-7b-chat

INT4-MIXED

32

140.4

82.5

12.12121212

deepseek-r1-distill-qwen-7b

INT4-MIXED

32

174.5

82.7

12.09189843

phi-3.5-vision-instruct

INT8-CW

1032

1541.4

83.1

12.03369434

qwen2-7b-instruct

INT4-MIXED

32

172.3

83.2

12.01923077

qwen2.5-7b-instruct

INT4-MIXED

32

172.5

83.3

12.00480192

llama-3.1-8b-instruct

INT4-MIXED

32

145.8

83.7

11.9474313

llama-3-8b-instruct

INT4-MIXED

32

142.8

83.8

11.93317422

qwen2.5-vl-7b-instruct

INT4-MIXED

32

134.9

84

11.9047619

qwen3_8b_draft_eagle3

INT4-MIXED

1024

1575.3

84

11.9047619

qwen2.5-7b-instruct-1m

INT4-MIXED

32

176.3

84.1

11.89060642

mistral-7b-instruct-v0.3

INT4-MIXED

1024

1314.6

84.2

11.87648456

qwen2-7b-instruct

INT4-MIXED

32

174.7

84.2

11.87648456

qwen2.5-7b-instruct

INT4-MIXED

32

179.5

84.2

11.87648456

deepseek-r1-distill-llama-8b

INT4-MIXED

35

129.4

84.5

11.83431953

mistral-7b-instruct-v0.2

INT4-MIXED

1024

1327.2

84.5

11.83431953

bloomz-7b1

INT4-MIXED

32

140.3

84.6

11.82033097

minicpm-v-2_6

INT4-MIXED

228

938.1

84.7

11.80637544

neural-chat-7b-v3-3

INT4-MIXED

1024

1490.2

84.9

11.77856302

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

1178.3

84.9

11.77856302

mistral-7b-instruct-v0.3

INT4-MIXED

1024

1495.4

85

11.76470588

qwen2.5-7b-instruct

INT4-MIXED

1024

1180.9

85

11.76470588

mistral-7b-instruct-v0.2

INT4-MIXED

1024

1488.2

85.2

11.7370892

qwen-7b-chat

INT4-MIXED

32

172.1

85.3

11.72332943

mistral-7b-instruct-v0.1

INT4-MIXED

1025

1597.5

85.4

11.70960187

qwen2.5-vl-7b-instruct

INT4-MIXED

32

140.5

85.4

11.70960187

minicpm-v-2_6

INT4-MIXED

228

940.8

85.5

11.69590643

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

1222.8

85.6

11.68224299

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

1248

85.7

11.66861144

qwen3-8b

INT4-MIXED

32

145.4

85.9

11.64144354

minicpm-o-2_6

INT4-MIXED

238

951.7

86

11.62790698

llama-3-8b-instruct

INT4-MIXED

32

175.4

86.1

11.61440186

llama-3-8b-instruct

INT4-MIXED

32

175.1

86.2

11.60092807

fara-7b

INT4-MIXED

32

218.4

86.3

11.58748552

llama-3.1-8b-instruct

INT4-MIXED

32

174.6

86.3

11.58748552

minicpm4-8b

INT4-MIXED

32

152.3

86.5

11.56069364

qwen3-4b

INT8-CW

1024

923.4

86.5

11.56069364

deepseek-r1-distill-qwen-7b

INT4-MIXED

1024

1428.8

86.8

11.52073733

qwen2.5-7b-instruct

INT4-MIXED

1024

1253.4

87.1

11.48105626

falcon-7b-instruct

INT4-MIXED

32

189.4

87.2

11.46788991

llama-3-8b-instruct

INT4-MIXED

32

178.4

87.2

11.46788991

qwen2-7b-instruct

INT4-MIXED

1024

1314.2

87.2

11.46788991

qwen2-7b-instruct-gptq

INT4-MIXED

32

175.8

87.2

11.46788991

afm-4.5b

INT8-CW

32

160.9

87.3

11.45475372

bloomz-7b1

INT4-MIXED

32

166.7

87.5

11.42857143

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

1382.2

87.6

11.41552511

minicpm-v-4_5

INT4-MIXED

217

950.8

87.8

11.38952164

qwen-7b-chat

INT4-MIXED

1024

1207.7

88

11.36363636

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

1301.8

88.1

11.3507378

qwen2.5-7b-instruct

INT4-MIXED

1024

1432.9

88.2

11.33786848

qwen2.5-7b-instruct-1m

INT4-MIXED

1024

1432.1

88.2

11.33786848

qwen2-7b-instruct

INT4-MIXED

1024

1435.6

88.3

11.32502831

qwen3-8b

INT4-MIXED

32

184.4

88.6

11.28668172

llama-3-8b-instruct

INT4-MIXED

1024

1227.8

89

11.23595506

llama-3.1-8b-instruct

INT4-MIXED

1024

1213.2

89.3

11.19820829

qwen2.5-vl-7b-instruct

INT4-MIXED

1024

1489.7

89.4

11.18568233

minicpm4-8b

INT4-MIXED

32

193.1

89.5

11.17318436

qwen3-8b

INT4-MIXED

32

185.6

89.9

11.12347052

bloomz-7b1

INT4-MIXED

1024

1615.8

90.1

11.09877913

deepseek-r1-distill-llama-8b

INT4-MIXED

1135

1492.6

90.2

11.0864745

minicpm3-4b

INT8-CW

32

168.5

90.4

11.0619469

fara-7b

INT4-MIXED

1024

1776.8

90.5

11.04972376

afm-4.5b

INT8-CW

1024

910.3

90.7

11.02535832

qwen-7b-chat

INT4-MIXED

1024

1307.4

90.7

11.02535832

zephyr-7b-beta

INT4-MIXED

32

180.8

90.7

11.02535832

llama-3-8b-instruct

INT4-MIXED

1024

1317.4

91.3

10.95290252

qwen2-7b-instruct-gptq

INT4-MIXED

1024

1275.8

91.3

10.95290252

llama-3-8b-instruct

INT4-MIXED

1024

1328

91.4

10.94091904

llama-3.1-8b-instruct

INT4-MIXED

1024

1316.4

91.7

10.90512541

qwen3-8b

INT4-MIXED

1024

1236

91.9

10.88139282

minicpm-v-4_5

INT4-MIXED

217

973.5

92

10.86956522

minicpm4-8b

INT4-MIXED

1024

1284.6

92.2

10.84598698

minicpm4-8b

INT4-MIXED

32

194.9

92.3

10.83423619

bloomz-7b1

INT4-MIXED

1024

2020.8

92.4

10.82251082

llama-3-8b-instruct

INT4-MIXED

1024

1502.9

92.4

10.82251082

falcon-7b-instruct

INT4-MIXED

1024

1519.8

92.7

10.78748652

llava-next-video-7b-hf

INT4-MIXED

2945

566.8

93.8

10.66098081

qwen3-8b

INT4-MIXED

1024

1332.5

94.4

10.59322034

qwen3.5-9b

INT4-MIXED

83

286.1

94.6

10.57082452

minicpm4-8b

INT4-MIXED

1024

1418.3

95.1

10.51524711

qwen3-8b

INT4-MIXED

1024

1545.3

95.7

10.44932079

qwen3.5-9b

INT4-MIXED

1024

1852.6

95.9

10.42752868

phi-4-multimodal-instruct

INT8-CW

578

1170.5

96

10.41666667

qwen3-vl-4b-instruct

INT8-CW

4937

2719.7

96.1

10.40582726

qwen3-vl-4b-instruct

INT8-CW

4907

2622

96.2

10.3950104

zephyr-7b-beta

INT4-MIXED

1024

1373.1

96.3

10.38421599

baichuan2-7b-chat

INT4-MIXED

32

190.6

96.4

10.37344398

phi-4-multimodal-instruct

INT8-CW

786

738.8

96.9

10.31991744

gemma-7b-it

INT4-MIXED

32

177.2

97.6

10.24590164

minicpm4-8b

INT4-MIXED

1024

1626.6

98

10.20408163

gemma-2b-it

FP16

32

115.2

98.4

10.16260163

phi-4-multimodal-instruct

INT8-CW

1362

2787.8

98.9

10.11122346

gemma-2b-it

FP16

1024

646

99.4

10.06036217

phi-4-multimodal-instruct

INT8-CW

1570

1402

99.9

10.01001001

All models listed here were tested with the following parameters:

  • Framework: PyTorch

  • Beam: 1

  • Batch size: 1