List your company

Speculative Decoding on Med

128.51Throughput (tokens/s)EvoSpec
88.282898.7264109.17119.6136Apr 17, 2026

Evaluation results

MethodLinks
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=2562026.04128.51
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=5122026.04128.23
EvoSpecStatic Vocabulary Size=32k, Dynamic Buffer Size (Dyna)=10242026.04127.8
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=10242026.04122.88
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=5122026.04118.46
EvoSpecStatic Vocabulary Size=16k, Dynamic Buffer Size (Dyna)=2562026.04116.21
FR-SpecStatic Vocabulary Size=32k2026.04114.08
Vanilla SD2026.04111.93
FR-SpecStatic Vocabulary Size=32k, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04111.35
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04110.85
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04110.47
EvoSpecStatic Vocabulary Size=32k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04109.7
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=10242026.04108.75
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04105.63
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04105.23
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=5122026.04105.13
EvoSpecStatic Vocabulary Size=16k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.04102.72
EvoSpecStatic Vocabulary Size=8k, Dynamic Buffer Size (Dyna)=2562026.04101.47
Standard SDDraft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.0498.58
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=1024, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.0495.09
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=512, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.0495.04
EvoSpecStatic Vocabulary Size=8k, Dynamic Vocabulary Size=256, Draft Model=Qwen3-1.7B, Target Model=Qwen3-4B, Architecture=Decoupled2026.0489.83