# vLLM v0.33.0 adds new models, faster speculative decoding, and PyTorch 2.14 support

> Original title: vLLM v0.33.0 is out! 🚀

- Source: vLLM
- Published: 2026-11-03T23:53:16.000Z
- Bad Influence: https://badinfluence.news/items/axyj9jodf3p5y1a9pkeaaw8b8
- Original: https://x.com/vllm_project/status/2117767095090053475

## Summary

vLLM v0.33.0 adds support for new models including Reflection Beam-1.0, Step-5 Preview, Nemotron Parse 3.5, and MiniCPM-OCR 1.0. The release reports 1.5–2x NVIDIA GB300 decode latency gains, 39–48% lower TTFT, and a 1.04–1.31x speedup from FLy lossy verify. It also moves to PyTorch 2.14 with CUDA 13.0-only wheels and ships a new AWS EFA image.

## Post

vLLM v0.33.0 is out! 🚀

927 commits · 389 contributors · 122 first-timers

🤖 New models: Reflection Beam-1.0, Step-5 Preview, Nemotron Parse 3.5, MiniCPM-OCR 1.0
🔥 PyTorch 2.14 — CUDA 13.0-only wheels, libtorch ABI-stable
🎲 Spec decoding: XPress refiner, FLy lossy verify (1.04–1.31x), DSpark 1373→3390 tok/s on GB300
⚡ NVIDIA decode: 1.5–2x latency decode on GB300, FA4 sparse MLA, 39–48% lower TTFT
🚀 New AWS EFA image + shorter vllm/vllm name
🐧 ROCm 10.0 default, MonoKernel 2.32x P50

Full notes ➡️ http://github.com/vllm-project/vllm/releases/tag/v0.33.0

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWPR8h-aIAAUbBp.jpg&mode=thumb&exp=1794009600&sig=bcec2f02c3d40416)