replicate AI Models

Tracked by OpenModelStats since Sep 15, 2026 · View on Hugging Face ↗

Tracked models
68
Downloads 30D
206.3K
Total downloads
206.3K
Total likes
1
Gained 7D
4,072

Most downloaded

  1. 1flash-attn3Sep 1, 2026202.6K
  2. 2vllm-flash-attn3Sep 1, 202610K
  3. 3punica-sgmvSep 1, 2026174
  4. 4yosoSep 1, 2026170
  5. 5flash-attn2Sep 1, 2026165
  6. 6flash-mlaSep 1, 2026162
  7. 7rotarySep 1, 2026158
  8. 8deformable-detrSep 1, 2026143
  9. 9rwkvSep 1, 2026134
  10. 10megablocksSep 1, 2026121

Fastest growing

  1. 1flash-attn3+3,248 in 7D202.6K
  2. 2vllm-flash-attn3+169 in 7D10K

Recently updated

  1. 1msaSep 21, 20260
  2. 2gemmSep 20, 20260
  3. 3deep-gemmSep 18, 20260
  4. 4yosoSep 1, 2026170
  5. 5vllm-flash-attn3Sep 1, 202610K
  6. 6triton_kernelsSep 1, 202626
  7. 7triton-scaled-mmSep 1, 20260
  8. 8triton-moeSep 1, 202611
  9. 9triton-layer-normSep 1, 20268
  10. 10triton-kernelsSep 1, 20268

All models

All tracked models published by replicate
ModelDL 30D7D
quantization-gptqreplicate/quantization-gptq44—
metal-flash-sdpareplicate/metal-flash-sdpa44—
sgl-flash-attn3replicate/sgl-flash-attn343—
layer_normreplicate/layer_norm43—
cv_utilsreplicate/cv_utils42+42
fp8-fbgemmreplicate/fp8-fbgemm38—
finegrained-fp8replicate/finegrained-fp836+36
trimul_gpumodereplicate/trimul_gpumode32—
moe-new-modelsreplicate/moe-new-models30—
bitsandbytes-mpsreplicate/bitsandbytes-mps30+30
quantization-bitsandbytesreplicate/quantization-bitsandbytes29—
triton_kernelsreplicate/triton_kernels26—
mlx-quantization-metal-kernelsreplicate/mlx-quantization-metal-kernels25—
liger-kernelsreplicate/liger-kernels23—
scattermoereplicate/scattermoe22—
mlx-rmsnormreplicate/mlx-rmsnorm21—
trimul-gpumodereplicate/trimul-gpumode19—
residual_rms_rocmreplicate/residual_rms_rocm19—
flash-attn-opsreplicate/flash-attn-ops16—
gpt-oss-triton-kernelsreplicate/gpt-oss-triton-kernels16—
triton-moereplicate/triton-moe11—
sonic-moereplicate/sonic-moe10—
aiter-ropereplicate/aiter-rope9+9
triton-layer-normreplicate/triton-layer-norm8—
triton-flash-attn-sinkreplicate/triton-flash-attn-sink8—