vLLM adds low/medium/high reasoning effort for Mistral Large 4, multi-image Pixtral
vllm_project · x · 2026-10-29
A new vLLM release adds low/medium/high reasoning effort levels (defaulting to high) for Mistral Large 4, and Pixtral vision now supports multiple images per prompt for multimodal workflows. SkyRL also ships day-0 RL training support for Mistral Large 4 with vLLM rollouts.
More from Infra
- Evo 2 genomic model now runs on vLLM: 7B fits 1-megabase context on a single H100 — anshulkundaje · 2026-10-30
- Meta's GenRecs Tokenization lifts ad conversions +0.35% with semantic vocabulary — Meta_Engineers · 2026-10-30
- Parallel Language Models paper: multi-token generation at AR quality, up to 5.7x speedup — aaron_lou · 2026-10-30
- Oracle investor day: 1GW AI site earns ~$80B contract value at ~28% ROI, 3.6-year payback — rohanpaul_ai · 2026-10-30
- Sam Altman: Local inference is a terrible allocation of scarce global compute — morqon · 2026-10-30
- AWS blueprint: deploy NVIDIA Nemotron 3.5 Super VL (120B MoE) on EKS with Dynamo and vLLM — AWS ML Blog · 2026-10-30