# vLLM v0.7.1 — Release & Upgrade Intelligence
- **Product:** vLLM (https://vllm.ai)
- **Version:** v0.7.1
- **Released:** Feb 01, 2025
- **Impact Score:** 5/10
- **Breaking Changes:** Not verified
- **Canonical Analysis:** https://changelogs.us/vllm/updates/v0.7.1/

## Executive Summary (TL;DR)
vLLM v0.7.1. Source highlight: This release features MLA optimization for Deepseek family of models. Compared to v0.7.0 released this Monday, we offer ~3x the generation throughput, ~10x the memory capacity for tokens, and horizontal context scalability with pipeline parallelism

## New Features
- Added a new design document for zero overhead prefix caching here (#12598)
- Add metrics and enhance logging for V1 engine (#12569, #12561, #12416, #12516, #12530, #12478)
- [V1][Metrics] Add initial Prometheus logger
- [Feature] [Spec decode]: Enable MLPSpeculator/Medusa and `prompt_logprobs` with ChunkedPrefill
- [FEATURE] Enables offline /score for embedding models
- [V1][Metrics] Add per-request prompt/generation_tokens histograms

## Improvements
- [V1] Improve Error Message for Unsupported Config
- [Doc] Improve installation signposting

## Bug Fixes
- [Bugfix] Fix missing seq_start_loc in xformers prefill metadata
- [Bugfix] Fix gpt2 GGUF inference
- Fix bad path in prometheus example
- [CI/Build] Fixed the xla nightly issue report in #12451
- [CI] fix pre-commit error
- [Doc] Fix typo for x86 CPU installation

## Upgrade Command
```bash
python -m pip install 'vllm==0.7.1'
```