Skip to main content
← Back to news

vLLM v0.33.0rc3: CUDA 13.0, new MoE models and speculative decoding on AMD

Tools & releases
Source:GitHub
927 commits and 389 contributors: vLLM moves to PyTorch 2.14 and Triton 3.8 with CUDA 13.0 only, ABI-stable wheels and an AWS EFA image. Beam-1.0 (501B MoE, 23B active) and Flint (40B, 2.5B active) land, and the XPress refiner with FLy lossy verification adds 1.04-1.31x decode on MI355X and B300.

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.