# vLLM v0.3.1 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.3.1
- **Released:** Feb 16, 2024
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.3.1/

## Executive Summary (TL;DR)
vLLM v0.3.1. Source highlight: This version fixes the following major bugs:

## New Features
- Add unit test for Mixtral MoE layer
- Refactor Prometheus and Add Request Level Metrics
- Add Internlm2
- Add one example to run batch inference distributed on Ray
- add requirement: triton >= 2.1.0
- Add fused top-K softmax kernel for MoE

## Bug Fixes
- Fixes assertion failure in prefix caching: the lora index mapping should respect `prefix_len`.
- fix some bugs about parameter description
- [Minor] Fix test_cache.py CI test failure
- Fix compile error when using rocm
- fix python 3.8 syntax
- fix length_penalty default value to 1.0

## Upgrade Command
```bash
python -m pip install 'vllm==0.3.1'
```