v0.20.2

vllm-project/vllmv0.20.2May 10, 2026by khluu

AI Summary

This is a small patch release that addresses critical stability issues in DeepSeek V4, gpt-oss, and Qwen3-VL models, specifically fixing hangs, KV cache allocation failures, and kernel capture problems.

Key Highlights

DeepSeek V4 sparse attention fixes for Hopper GPUs and CUDA graph capture
Fixed 'failure to allocate KV blocks' error in DeepSeek V4
gpt-oss MXFP4 support under torch.compile
Qwen3-VL boundary check fix for heavy load scenarios

New Features

MXFP4 support under torch.compile for gpt-oss
DeepSeek V4 sparse attention kernel improvements

Full Release Notes

# vLLM v0.20.2

## Highlights
This release features 6 commits from 6 contributors (0 new)!

This is a small patch release with bug fixes for DeepSeek V4, gpt-oss, and Qwen3-VL

### Bug Fixes
* **DeepSeek V4 sparse attention**: Re-enable the persistent topk path on Hopper and ensure the memset kernel runs at CUDA graph capture time regardless of `max_seq_len`, fixing the MTP=1 hang on DeepSeek V4 (#41665, revert of #41605).
* **DeepSeek V4 KV cache**: Fixed a "failure to allocate KV blocks" error in the V1 engine KV cache manager (#41282).
* **gpt-oss MXFP4 + torch.compile**: Plumbed `hidden_dim_unpadded` through the `moe_forward` fake op so MXFP4 works under `torch.compile` on v0.20.x (#42002, backport of #41646).
* **Qwen3-VL**: Removed an invalid deepstack boundary check that could fail under heavy load (#40932).

## Contributors
@ywang96, @zyongye, @stecasta, @wzhao18, @Isotr0py, @khluu