# vLLM v0.13.0 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.13.0
- **Released:** Dec 19, 2025
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.13.0/

## Executive Summary (TL;DR)
vLLM v0.13.0. Source highlight: vLLM v0.13.0 Release Notes Highlights

## Breaking Changes & Deprecations
- Breaking Changes**: This release includes deprecation removals, PassConfig flag renames, and attention configuration changes from environment variables to CLI arguments. Please review the breaking changes section carefully before upgrading.
- Breaking Changes & Deprecations 1. **PassConfig flags renamed** per RFC #27995 (#29646) 2. **Attention env vars → CLI args**: `VLLM_ATTENTION_BACKEND` replaced with `--attention-backend` (#26315)

## Improvements
- **Prefix caching**: xxHash high-performance hash option (#29163).
- **Speculative decoding**: Medusa GPU-CPU sync avoidance (#29723), async spec-decode improvements (#29624).
- **Whisper**: Major performance improvements - V1 is now faster than V0 (~3x speedup vs v0.12.0). Encoder batching (#29421), `FULL_DECODE_ONLY` CUDA graph (#30072), CPU backend support (#30062).
- **Performance**: Fused blockwise quant RMS norm (#27883), MoE LoRA loading reduction (#30243), encoder cache optimization (#30475), CPU KV offloading streams (#29013).
- **NVIDIA Blackwell Ultra**: SM103 (GB300) support with CUDA 13 (#30484).
- **DeepSeek optimizations** (benchmarked on DeepSeek-V3.1):

## Bug Fixes
- **Model enhancements**: Qwen3-VL embeddings support (#30037), Qwen3-VL EVS (Efficient Video Sampling) (#29752), DeepSeek V3.2 proper `drop_thinking` logic (#30490), DeepSeek V3.2 top-k fix (#27568).
- **Anthropic API**: Streaming fixes (#29971, #30266).
- NVSHMEM 3.3.24 + CUDA 13 fix (#30149).

## Upgrade Command
```bash
python -m pip install 'vllm==0.13.0'
```