# vLLM v0.6.3 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.6.3
- **Released:** Oct 14, 2024
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.6.3/

## Executive Summary (TL;DR)
vLLM v0.6.3. Source highlight: Text: Granite MoE (#8206), Mamba (#6484, #8533)

## New Features
- Add Gemma2 embedding model (#9004)
- Add support for Llama 3.1 and 3.2 tool use (#8343)
- New compatibility matrix for mutual exclusive features (#8512)
- Add async output processor for xpu (#8897)
- Add on-device sampling support for Neuron (#8746)
- [Docs] Add README to the build docker image

## Improvements
- Out of tree support enhancements: Explicit interface for vLLM models and support OOT embedding models (#9108)
- Progress in vLLM's refactoring to a core core:
- Spec decode removing batch expansion (#8839, #9298).
- We have made block manager V2 the default. This is an internal refactoring for cleaner and more tested code path (#8678).
- Moving beam search from the core to the API level (#9105, #9087, #9117, #8928)
- Move guided decoding params into sampling params (#8252)

## Bug Fixes
- [CI/Build] Fix missing ci dependencies
- [Bugfix] Include encoder prompts len to non-stream api usage response
- [Bugfix] Fix print_warning_once's line info
- fix validation: Only set tool_choice `auto` if at least one tool is provided
- [Bugfix] Fixup advance_step.cu warning
- [BugFix] Fix test breakages from transformers 4.45 upgrade

## Upgrade Command
```bash
python -m pip install 'vllm==0.6.3'
```