vLLM releases v0.32.0rc3: fixes the Mamba2 chunk scan kernel NaN on sm90
Tools & releasesSource:GitHub
vLLM's v0.32.0rc3 release candidate fixes the NaN output of the Mamba2 chunk-scan kernel on sm90. Relevant for anyone serving hybrid state-space (Mamba2) models on Blackwell hardware: it's the build that also serves decision models like JEV-35B.
Related articles
Running LLM Models Locally: From First Install to First Working Model
From zero to a first answering model: what you actually need, Ollama step by step, how to pick a model for your memory, expected speeds per hardware, and the typical problems fixed.
When a Big Model Isn't Needed: Choosing by Size
Memory, speed and latency: how to pick the right model size for your hardware, and when the small model genuinely wins.
Hardware for Local AI: What You Need in 2026
The rule (weights + KV cache), real October 2026 prices for NVIDIA GPUs, Apple Silicon and CPU, the 36-month cost math and the choice by budget.
Get the next local-AI briefing
Once a week, the top selection from iaLocale: articles, news and resources on local AI, and more.
After signing up you receive a confirmation email (double opt-in). Without confirmation we send nothing. No spam, we promise.