List your company

Training Throughput on Open-PerfectBlend

0.52Step Time (s)SpecForge
0.09282.97645.868.7436Mar 19, 2026

Evaluation results

MethodLinks
SpecForgeTarget Model=Qwen3-30B-A3B, Target Model Parallelism=TP=4, Draft Model Parallelism=ZeRO 2, Max Batch Size=16, Seq Length=40962026.030.52126,030.84.31
EAGLETarget Model=Llama3.1-8B, Target Model Parallelism=ZeRO 2, Draft Model Parallelism=ZeRO 2, Max Batch Size=16, Seq Length=40962026.031.0463,015.41
EAGLETarget Model=Qwen3-30B-A3B, Target Model Parallelism=ZeRO 2, Draft Model Parallelism=ZeRO 2, Max Batch Size=8, Seq Length=40962026.031.1229,257.11
SpecForgeTarget Model=Qwen3-235B-A22B, Target Model Parallelism=TP=8, Draft Model Parallelism=ZeRO 2, Max Batch Size=8, Seq Length=40962026.031.6220,227.29.99
SpecForgeTarget Model=Llama3.1-8B, Target Model Parallelism=TP=1, Draft Model Parallelism=ZeRO 2, Max Batch Size=64, Seq Length=40962026.032.07126,639.62.01
EAGLETarget Model=Llama3.3-70B, Target Model Parallelism=ZeRO 3, Draft Model Parallelism=ZeRO 3, Max Batch Size=8, Seq Length=40962026.032.2114,827.11
SpecForgeTarget Model=Llama3.3-70B, Target Model Parallelism=TP=4, Draft Model Parallelism=ZeRO 2, Max Batch Size=16, Seq Length=40962026.033.1820,608.81.39
EAGLETarget Model=Qwen3-30B-A3B, Target Model Parallelism=ZeRO 3, Draft Model Parallelism=ZeRO 3, Max Batch Size=8, Seq Length=40962026.035.076,463.10.2
EAGLETarget Model=Qwen3-235B-A22B, Target Model Parallelism=ZeRO 3, Draft Model Parallelism=ZeRO 3, Max Batch Size=8, Seq Length=40962026.0311.22,025.71