Zyphra ZYFA: FP8 FlashAttention for AMD MI355X, Up to 1.76x Faster

modelsMarkTechPostNov 6, 2026

Zyphra released ZYFA, open-source FP8 and BF16 prefill attention kernels for AMD Instinct MI355X. On real model activations, its FP8 kernels run 1.16-1.76x faster than AMD’s AITER, with 62-78% lower error at head dimension 128. The post Zyphra ZYFA: FP8 Fla...

Impact: LowConfidence: LowNovelty: Incremental
modelsmarktechpost

See the whole story

Read the original on MarkTechPost

Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original source is authoritative.