vLLM has released version 0.32.0rc3, which includes a bug fix for the Mamba2 chunk scan kernel on SM 90 hardware. This release addresses an issue that could lead to NaN outputs. The fix was contributed by users Vadim Gimpelson and vadiklyutiy, with a co-authored commit from Claude. AI
IMPACT Minor bug fix for vLLM's Mamba2 kernel improves stability on specific hardware.
RANK_REASON This is a minor bugfix release for an open-source inference library, not a frontier model release or significant industry event.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →