# EricLBuehler/mistral.rs

Fast, flexible LLM inference

Repository: https://github.com/EricLBuehler/mistral.rs
Canonical: https://ross.abutalabs.com/products/mistralrs
Language: Rust
License: MIT
License Family: permissive
Topics: llm, rust, uqff
Last push: 2026-08-25T21:46:44+00:00

## Health v2 (maintenance only)
Score: 92/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 98, longevity 65
- inputs: {"age_days": 919, "days_push": 8, "days_rel": 13, "gap_med": 0.0, "n_releases_24m": 33}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7628, forks 685 (observed 2026-08-28T04:10:02.087196+00:00)

## What it is
mistral.rs is a fast and flexible LLM inference engine written in Rust, supporting many model families with quantization (ISQ/UQFF, GGUF), paged attention, and multimodal inputs. It exposes OpenAI- and Anthropic-compatible HTTP APIs and provides Rust and Python SDKs for local or server-based inference.

## Use cases
- serve local llms with an openai-compatible api
- run quantized gguf models on gpu
- self-host a chat completion server
- run multimodal llm inference with images and audio
- build agents with local llm tool calling
- benchmark llm inference performance in rust
- expose anthropic messages api from a local model

## When to choose
- you need high-performance local LLM serving in Rust with OpenAI-compatible endpoints
- you want quantized inference (GGUF, ISQ, UQFF) with broad model support
- you need multimodal inputs, tool calling, or agentic features on self-hosted models

## When to avoid
- you only need a simple Python-only inference stack without a serving layer
- you require the largest ecosystem of community integrations available in mainstream runtimes
- you need CPU-only inference on very constrained hardware

## Facets
- artifact type: service
- maturity: active
- function: llm-inference, http-server, api-framework, rag, agent-framework, gpu-computing
- domain: large-language-models, artificial-intelligence, developer-tools, self-hosted
- platform: rust, python, windows, cli
- tags: openai-compatible-api, gguf, quantization, multimodal, paged-attention, anthropic-api, serving, linux, macos, cuda, web-server

## Member repositories
- EricLBuehler/mistral.rs (main) score 92

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:02.087196+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:36:47.785737+00:00, confidence not recorded.
  - readme: https://github.com/EricLBuehler/mistral.rs (fetched 2026-08-28T04:10:02.087196+00:00, sha 1f0653e976cb)
- Data as of 2026-08-30T08:39:29.467469+00:00.