---
title: README
emoji: 🐆
colorFrom: yellow
colorTo: gray
sdk: static
pinned: false
---

## Run it on your Mac

Rapid-MLX is a local OpenAI- and Anthropic-compatible inference server for Apple Silicon.

```bash
curl -fsSL https://rapidmlx.com/install.sh | bash
rapid-mlx serve ling-3.0-tiny-4bit
curl -s http://localhost:8000/v1/chat/completions -H 'Content-Type: application/json' -d '{"model":"default","messages":[{"role":"user","content":"Say hello"}]}'
```

Use it from OpenCode / Claude Code / Codex: `rapid-mlx agents <name> --setup` · [Agent setup](https://rapidmlx.com/docs/agents)

[Download the Desktop app](https://rapidmlx.com/download) · The example needs at least 8 GB unified memory; leave extra headroom for long contexts.

[See the measured comparison](https://rapidmlx.com/compare)

# Rapid-MLX

**Rapid-MLX** is an open-source (Apache 2.0) OpenAI- and Anthropic-compatible LLM inference server for Apple Silicon, built on MLX, focused on reliable tool calling for coding agents.

**Up to 4× faster than Apple's MLX (mlx-lm), 1.5× on a typical task — same model, same weights.**
Qwen3.5-9B 4-bit on a Mac mini M4 Pro, greedy decoding, Rapid-MLX 0.15.6 vs mlx-lm 0.32.0 at defaults, measured 2026-10-04. Per-task results, method and raw data: [rapidmlx.com/compare/mlx-lm](https://rapidmlx.com/compare/mlx-lm)

```bash
curl -fsSL https://rapidmlx.com/install.sh | bash
rapid-mlx serve qwen3.5-9b-4bit        # OpenAI-compatible API on http://127.0.0.1:8000/v1
```

Any MLX model on the Hub runs the same way: `rapid-mlx serve <org/repo>` ([bring your own model](https://rapidmlx.com/docs/byom)).

- Website and docs: [rapidmlx.com](https://rapidmlx.com)
- Source: [github.com/raullenchai/Rapid-MLX](https://github.com/raullenchai/Rapid-MLX)
- Which model fits your Mac: [community leaderboard](https://rapidmlx.com/leaderboard)