Qwen3.6 Hits 65 tok/s on Dual B60s
QuixiAI · x · 2026-07-10
A user ran nvidia/Qwen3.6-35B-A3B-NVFP4 on dual B60s, optimized with a custom SYCL kernel. Tests show a context length of up to 128k and an inference speed of 65 tok/s, with DFlash disabled.
More from Infra
- Strangeworks launches Aura to turn enterprise ops into production optimization systems — whurley · 2026-07-22
- Graph workload 854.graph500 enters SPEC CPU 2026 as a new CPU benchmark — Prof_DavidBader · 2026-07-22
- HilbertRaum open-sources a fully local AI chat and document analysis app for private use — Vladowski · 2026-07-22
- Hybrid and local inference are emerging as a response to AI energy and token costs — dmitry140 · 2026-07-22
- NVIDIA details Vera CPU with 2x performance claims and a 22,000-core rack — ryanshrout · 2026-07-22
- NVIDIA says Vera Rubin NVL72 delivers 10x more tokens per megawatt than Blackwell — nvidia · 2026-07-22