# Rapid-MLX

Runs open models locally on Apple silicon and serves them through an OpenAI- and Anthropic-compatible API.

HTML: https://softbests.com/software/rapid-mlx
Markdown: https://softbests.com/software/rapid-mlx.md
Website: https://rapidmlx.com

- Category: [Developer utilities](https://softbests.com/category/developer-tools)
- Pricing: Open source
- Platforms: macOS, CLI
- Underrated

## Best for

Mac users who want Claude Code, Codex or Cursor running against a local model instead of a paid API.

## Why it deserves attention

Rapid-MLX installs with one command, downloads from a catalogue of 240-plus open models, and exposes both OpenAI- and Anthropic-style endpoints, so existing agent tools work by changing a base URL. It is built on Apple's MLX and claims up to four times the throughput of stock MLX on the same hardware, with text, vision, speech and video models included. Everything runs on-device, so prompts and files never leave the Mac, and it is Apache-2.0 and free. The limits are real: Apple silicon and macOS 14+ only, a server rather than a GUI, and model quality is bounded by the memory your machine has. The speed figures come from the project's own benchmarks, so measure on your own hardware.

## Can replace

- [Ollama](https://softbests.com/software/ollama)

## Related and alternatives

- [LM Studio](https://softbests.com/software/lm-studio)
- [Claude Code](https://softbests.com/software/claude-code)