# vLLM v0.2.7 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.2.7
- **Released:** Jan 04, 2024
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.2.7/

## Executive Summary (TL;DR)
vLLM v0.2.7. Source highlight: Up to 70% throughput improvement for distributed inference by removing serialization/deserialization overheads

## New Features
- Add SSL arguments to API servers
- [Minor] Add warning on CUDA graph memory usage
- Added DeciLM-7b and DeciLM-7b-instruct
- Add "About" Heading to README.md
- Add support GPT-NeoX Models without attention biases

## Improvements
- Up to 70% throughput improvement for distributed inference by removing serialization/deserialization overheads

## Bug Fixes
- Fix tensor parallelism support for Mixtral + GPTQ/AWQ
- Minor fix for gpu-memory-utilization description
- [BugFix] Raise error when max_model_len is larger than KV cache size
- [BugFix] Fix RoPE kernel on long sequences
- typo fix
- [ROCm] Fixes for GPTQ on ROCm

## Upgrade Command
```bash
python -m pip install 'vllm==0.2.7'
```