Description
Native Zig inference server for Apple Silicon. Runs any LLM — MLX and GGUF — faster than LM Studio. OpenAI- and Anthropic-compatible (Claude Code) APIs. No Python. Drop-in alternative to LM Studio, Ollama, and llama.cpp on macOS. Gemma 4, Qwen 3.8, Llama 3, Mistral, DeepSeek V4 Flash — local, MIT-licensed.…
Ratings & Reviews
No ratings yet
Be the first to rate mlx-serve.
People also visit
mlx-serve FAQ
What is mlx-serve?
Native Zig inference server for Apple Silicon. Runs any LLM — MLX and GGUF — faster than LM Studio. OpenAI- and Anthropic-compatible (Claude Code) APIs. No Python. Drop-in alternative to LM Studio, Ollama, and llama.cpp on macOS. Gemma 4, Qwen 3.8, Llama 3, Mistral, DeepSeek V4 Flash — local, MIT-licensed.
Is mlx-serve free?
Yes. mlx-serve is free to use.
Do I need to download mlx-serve?
No. mlx-serve runs in your browser at mlxserve.com, so there's nothing to install.
Who makes mlx-serve?
mlx-serve is made by ddalcu.
What are the best mlx-serve alternatives?
Popular alternatives to mlx-serve include Rapid-MLX, vllm-mlx and OpenTrade. See all mlx-serve alternatives