Open-source XeMaya engine runs DeepSeek V4.1 on Intel Arc GPUs

Developer Jumbomuffin1312 released XeMaya, an open-source inference engine forked from XeStrata that uses SYCL, Level Zero, and Intel XMX acceleration to natively run DeepSeek V4.1 on Intel Arc GPUs, reaching 21.7 tok/s with 88GB of VRAM across three cards.

2026-11-11 ~ 2026-11-11 · 2 related posts

• XeMaya alpha: open-source Intel Arc inference engine runs DeepSeek V4.1 on 88GB of Arc GPUs — Jumbomuffin1312 · 2026-11-11

1 near-duplicate retellings: Jumbomuffin1312