List your company

Speculative Decoding on Law

132.69Throughput (tokens/s)EvoSpec
97.4444106.5947115.745124.8953Apr 17, 2026

Evaluation results

MethodLinks
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=2562026.04132.69
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=5122026.04132.51
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=10242026.04132.23
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=10242026.04128.41
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=5122026.04124.33
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=2562026.04119.54
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04116.27
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04115.55
FR-SpecStatic Vocabulary Size=32k2026.04114.61
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04114.27
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04113.64
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04113.29
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04112.58
FR-SpecStatic Vocabulary Size=32k, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04112.54
Vanilla SD2026.04112.2
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=10242026.04110.12
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=5122026.04108.57
Standard SDDraft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04107.12
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04106.29
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=2562026.04105.19
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04104.64
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.0498.8