Fast, flexible LLM inference.

| Documentation | Rust SDK | Python SDK | Discord |

Latest

• Gemma 4: Full multimodal: text, image, video, and audio input. Guide | Video setup
• MXFP4 ISQ quantization: MXFP4 with optimized decode kernels for faster, smaller models. Quantization docs
• Qwen 3.5 model family: Support for the Qwen 3.5 series including vision. Guide

Why mistral.rs?

• Any Hugging Face model, zero config: Just mistralrs run -m user/model.
• True multimodality: Text, vision, video, and audio, speech generation, image generation, and embeddings in one engine.
• Full quantization control: Choose the precise quantization you want to use, or make your own UQFF with mistralrs quantize.
• Built-in web UI: mistralrs serve --ui gives you a web interface instantly.
• Hardware-aware: mistralrs tune benchmarks your system and picks optimal quantization + device mapping.
• Flexible SDKs: Python package and Rust crate to build your projects.
• Agentic features — server-side tool loop, web search, MCP client, and HTTP tool dispatch

Quick Start

Install

Linux/macOS:

curl --proto '=https' --tlsv1.2 -sSf https://raw.githubusercontent.com/EricLBuehler/mistral.rs/master/install.sh | sh

Windows (PowerShell):

irm https://raw.githubusercontent.com/EricLBuehler/mistral.rs/master/install.ps1 | iex

Manual installation & other platforms

Run Your First Model

# Interactive chat
mistralrs run -m Qwen/Qwen3-4B

# One-shot prompt (no interactive session)
mistralrs run -m Qwen/Qwen3-4B -i "What is the capital of France?"

# One-shot with an image
mistralrs run -m google/gemma-4-E4B-it --image photo.jpg -i "Describe this image"

# Or start a server with web UI
mistralrs serve --ui -m google/gemma-4-E4B-it

Then visit http://localhost:1234/ui for the web chat interface.

The mistralrs CLI

The CLI is designed to be zero-config: just point it at a model and go.

• Auto-detection: Automatically detects model architecture, quantization format, and chat template
• All-in-one: Single binary for chat, server, benchmarks, and web UI (run, serve, bench)
• Hardware tuning: Run mistralrs tune to automatically benchmark and configure optimal settings for your hardware
• Format-agnostic: Works with Hugging Face models, GGUF files, and UQFF quantizations seamlessly

# Auto-tune for your hardware and emit a config file
mistralrs tune -m Qwen/Qwen3-4B --emit-config config.toml

# Run using the generated config
mistralrs from-config -f config.toml

# Diagnose system issues (CUDA, Metal, HuggingFace connectivity)
mistralrs doctor

Full CLI documentation

Web Chat Demo

What Makes It Fast

Performance

• Continuous batching support by default on all devices.
• CUDA with FlashAttention V2/V3, Metal, multi-GPU tensor parallelism
• PagedAttention for high throughput continuous batching on CUDA or Apple Silicon, prefix caching (including multimodal)

Quantization (full docs)

• In-situ quantization (ISQ) of any Hugging Face model
• GGUF (2-8 bit), GPTQ, AWQ, HQQ, FP8, BNB support
• ⭐ Per-layer topology: Fine-tune quantization per layer for optimal quality/speed
• ⭐ Auto-select fastest quant method for your hardware

Flexibility

• LoRA & X-LoRA with weight merging
• AnyMoE: Create mixture-of-experts on any base model
• Multiple models: Load/unload at runtime

Agentic Features

• Integrated tool calling with grammar enforcement and strict schema mode
• ⭐ Server-side agentic loop: auto-execute tools and feed results back
• ⭐ Tool dispatch URL: POST tool calls to your own endpoint
• ⭐ Web search integration with embedding-based ranking
• ⭐ MCP client: Connect to external tools via Process, HTTP, or WebSocket
• Python/Rust tool callbacks for custom execution

Full feature documentation

Supported Models

Request a new model | Full compatibility tables

Python SDK

pip install mistralrs  # or mistralrs-cuda, mistralrs-metal, mistralrs-mkl, mistralrs-accelerate

from mistralrs import Runner, Which, ChatCompletionRequest

runner = Runner(
    which=Which.Plain(model_id="Qwen/Qwen3-4B"),
    in_situ_quant="4",
)

res = runner.send_chat_completion_request(
    ChatCompletionRequest(
        model="default",
        messages=[{"role": "user", "content": "Hello!"}],
        max_tokens=256,
    )
)
print(res.choices[0].message.content)

Python SDK | Installation | Examples | Cookbook

Rust SDK

cargo add mistralrs

use anyhow::Result;
use mistralrs::{IsqType, TextMessageRole, TextMessages, MultimodalModelBuilder};

#[tokio::main]
async fn main() -> Result<()> {
    let model = MultimodalModelBuilder::new("google/gemma-4-E4B-it")
        .with_isq(IsqType::Q4K)
        .with_logging()
        .build()
        .await?;

    let messages = TextMessages::new().add_message(
        TextMessageRole::User,
        "Hello!",
    );

    let response = model.send_chat_request(messages).await?;

    println!("{:?}", response.choices[0].message.content);

    Ok(())
}

API Docs | Crate | Examples

Docker

For quick containerized deployment:

docker pull ghcr.io/ericlbuehler/mistral.rs:latest
docker run --gpus all -p 1234:1234 ghcr.io/ericlbuehler/mistral.rs:latest \
  serve -m Qwen/Qwen3-4B

Docker images

For production use, we recommend installing the CLI directly for maximum flexibility.

Documentation

For complete documentation, see the Documentation.

Quick Links:

• CLI Reference - All commands and options
• HTTP API - OpenAI-compatible endpoints
• Quantization - ISQ, GGUF, GPTQ, and more
• Device Mapping - Multi-GPU and CPU offloading
• MCP Integration - MCP integration documentation
• Troubleshooting - Common issues and solutions
• Configuration - Environment variables for configuration

Contributing

Contributions welcome! Please open an issue to discuss new features or report bugs. If you want to add a new model, please contact us via an issue and we can coordinate.

Credits

This project would not be possible without the excellent work at Candle. Thank you to all contributors!

mistral.rs is not affiliated with Mistral AI.

Back to Top