List your company

LLM Decoding on Llama 70B 3.1

3,119.55ThroughputDeepFusionKernel
-121.09720.231,561.552,402.87Feb 12, 2026

Evaluation results

MethodLinks
DeepFusionKernelBatch size=64, GPU=H100, Output length=10242026.023,119.553.4
SGLangBatch size=64, GPU=H100, Output length=10242026.023,016.4—
DeepFusionKernelBatch size=32, GPU=H100, Output length=10242026.021,635.945.5
SGLangBatch size=32, GPU=H100, Output length=10242026.021,551.33—
DeepFusionKernelBatch size=16, GPU=H100, Output length=10242026.02914.624.2
SGLangBatch size=16, GPU=H100, Output length=10242026.02878.08—
DeepFusionKernelBatch size=16, Output len=1024, Kernel scheduler=true2026.02493.6—
DeepFusionKernelBatch size=16, Output len=4096, Kernel scheduler=true2026.02478.66—
DeepFusionKernelBatch size=8, GPU=H100, Output length=10242026.02470.634.2
vLLMBatch size=16, Output len=1024, Kernel scheduler=true2026.02462.01—
SGLangBatch size=16, Output len=1024, Kernel scheduler=true2026.02453.75—
SGLangBatch size=16, Output len=4096, Kernel scheduler=true2026.02452.6—
SGLangBatch size=8, GPU=H100, Output length=10242026.02451.64—
vLLMBatch size=16, Output len=4096, Kernel scheduler=true2026.02448.93—
DeepFusionKernelBatch size=16, Output len=16384, Kernel scheduler=true2026.02416.19—
SGLangBatch size=16, Output len=16384, Kernel scheduler=true2026.02396.7—
vLLMBatch size=16, Output len=16384, Kernel scheduler=true2026.02390.94—
DeepFusionKernelBatch size=4, GPU=H100, Output length=10242026.02237.053.6
SGLangBatch size=4, GPU=H100, Output length=10242026.02228.9—
DeepFusionKernelBatch size=4, Output len=1024, Kernel scheduler=true2026.02140.15—
DeepFusionKernelBatch size=4, Output len=4096, Kernel scheduler=true2026.02138.16—
SGLangBatch size=4, Output len=1024, Kernel scheduler=true2026.02135.76—
SGLangBatch size=4, Output len=4096, Kernel scheduler=true2026.02132.66—
DeepFusionKernelBatch size=4, Output len=16384, Kernel scheduler=true2026.02131.75—
TorchBatch size=64, GPU=H100, Output length=10242026.02129.81—
SGLangBatch size=4, Output len=16384, Kernel scheduler=true2026.02126.81—
vLLMBatch size=4, Output len=1024, Kernel scheduler=true2026.02126.1—
vLLMBatch size=4, Output len=4096, Kernel scheduler=true2026.02125.1—
DeepFusionKernelBatch size=2, GPU=H100, Output length=10242026.02121.1313.2
vLLMBatch size=4, Output len=16384, Kernel scheduler=true2026.02119.3—
TorchBatch size=32, GPU=H100, Output length=10242026.02118.56—
SGLangBatch size=2, GPU=H100, Output length=10242026.02107.05—
DeepFusionKernelBatch size=1, GPU=H100, Output length=10242026.0262.223.4
SGLangBatch size=1, GPU=H100, Output length=10242026.0260.2—
TorchBatch size=16, GPU=H100, Output length=10242026.0252.37—
DeepFusionKernelBatch size=1, Output len=1024, Kernel scheduler=true2026.0237.57—
DeepFusionKernelBatch size=1, Output len=4096, Kernel scheduler=true2026.0237.5—
DeepFusionKernelBatch size=1, Output len=16384, Kernel scheduler=true2026.0236.79—
SGLangBatch size=1, Output len=1024, Kernel scheduler=true2026.0235.68—
SGLangBatch size=1, Output len=4096, Kernel scheduler=true2026.0235.53—
SGLangBatch size=1, Output len=16384, Kernel scheduler=true2026.0234.97—
TorchBatch size=8, GPU=H100, Output length=10242026.0234.3—
vLLMBatch size=1, Output len=4096, Kernel scheduler=true2026.0233.64—
vLLMBatch size=1, Output len=1024, Kernel scheduler=true2026.0233.47—
vLLMBatch size=1, Output len=16384, Kernel scheduler=true2026.0232.93—
TorchBatch size=4, GPU=H100, Output length=10242026.0218.11—
TorchBatch size=2, GPU=H100, Output length=10242026.028.39—
TorchBatch size=1, GPU=H100, Output length=10242026.023.55—