# vLLM v0.2.1 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.2.1
- **Released:** Oct 16, 2023
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.2.1/

## Executive Summary (TL;DR)
vLLM v0.2.1. Source highlight: PagedAttention V2 kernel: Up to 20% end-to-end latency reduction

## New Features
- Added `dtype` arg to benchmarks
- add support for tokenizer revision
- Add blacklist for model checkpoint

## Improvements
- Improve detokenization performance

## Bug Fixes
- fix vulnerable memory modification to gpu shared memory
- [Minor] Fix type annotations
- [FIX] Explain why the finished_reason of ignored sequences are length
- [Minor] Fix comment in mistral.py
- Fix overflow in awq kernel
- Fix the issue for AquilaChat2-* models

## Upgrade Command
```bash
python -m pip install 'vllm==0.2.1'
```