Eagle-4 Released: Faster than DFlash2 & DSpark
retained text
Insane numbers, so far only Qwen3.8-27B is supported but wow those are some good numbers. Will watch closely when this lands in llama.cpp/vLLM \### Qwen 3.8 27B Benchmark Results |**Method**|**Speedup over Vanilla**| |:-|:-| |MTP|2.22×| |DSpark|2.57×| |DFlash 2|2.80×| |EAGLE-4|3.59×|
document fetched 2026-10-21 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1xb6zds/eagle4_released_faster_than_dflash2_dspark