Models

123 models with kernel evidence · 6 serving models

129 models

Model

Operations

Kernel runs

GPUs

Serving runs

19

4,432

3

—

14

1,886

1

—

18

1,445

1

—

21

1,285

2

—

14

570

1

—

qwen3-vl-30b-a3b-instruct-fp8

7

253

1

—

9

212

1

—

1

202

4

—

parakeet-tdt-0-6b-v3

6

192

1

—

wan2-2-t2v-a14b

6

183

1

—

nvidia-nemotron-3-nano-30b-a3b-fp8

4

148

1

—

4

145

1

—

qwen3-omni-30b-a3b-instruct

5

144

1

—

sam-hq-vit-base

5

137

1

—

3

133

1

—

seed-seed-oss-36b-instruct

1

128

1

—

3

121

1

—

deepseek-r1-0528

4

119

1

—

qwen3-next-80b-a3b-thinking-nvfp4

4

116

1

—

nvidia-nemotron-3-nano-30b-a3b-nvfp4

4

112

1

—

ai21-jamba-reasoning-3b

1

110

1

—

qwen-image-edit-2509

3

98

1

—

hunyuanimage-3-0

3

97

1

—

stable-diffusion-3-5-medium

3

95

1

—

qwen3-14b-nvfp4

3

93

1

—

3

92

1

—

4

91

1

—

5

91

1

—

convnextv2-base-22k-224

3

90

1

—

3

89

1

—

Model relevance is workload provenance declared by the imported sources, not a completeness claim; kernel and serving counts are separate corpora and never one ranking.Sources and limitations JSON