SYLIN

EAGLE-3

by SafeAILab · Technique · Text & code

EAGLE-3 mostly below 1.2x on Qwen3-8B (below baseline on MATH500); up to 2.33x on Kimi-K2.5 and 2.09x on MiniMax-M3

Alternatives

DFlashOn MI300X/MI355X in vLLM, DFlash reached up to 2.87x throughput on Gemma 4 26B-A4B versus 2.11–2.27x for EAGLE-3