# vLLM v0.4.1 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.4.1
- **Released:** Apr 24, 2024
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.4.1/

## Executive Summary (TL;DR)
vLLM v0.4.1. Source highlight: Support and enhance CommandR+ (#3829), minicpm (#3893), Meta Llama 3 (#4175, #4182), Mixtral 8x22b (#4073, #4002)

## New Features
- Add option for using LM Format Enforcer for guided decoding (#3868)
- Add option for optionally initialize tokenizer and detokenizer (#3748)
- Add option for load model using `tensorizer` (#3476)
- Intel CPU inference backend is added (#3993, #3634)
- [Misc] Add support for new autogptq checkpoint_format
- [Hardware][Intel] Add CPU inference backend

## Improvements
- Support and enhance CommandR+ (#3829), minicpm (#3893), Meta Llama 3 (#4175, #4182), Mixtral 8x22b (#4073, #4002)
- AMD backend is enhanced with Triton kernel and e4m3fn KV cache (#3643, #3290)
- [Kernel] Layernorm performance optimization
- [Core] improve robustness of pynccl
- [Core][Distributed] improve logging for init dist

## Bug Fixes
- [Misc] Minor fixes in requirements.txt
- [Misc] Fix Benchmark TTFT Calculation for Chat Completions
- [HotFix] [CI/Build] Minor fix for CPU backend CI
- [Frontend][Bugfix] allow using the default middleware with a root path
- [Doc] Fix vLLMEngine Doc Page
- [CI/Build] fix TORCH_CUDA_ARCH_LIST in wheel build

## Upgrade Command
```bash
python -m pip install 'vllm==0.4.1'
```