Rapid-MLX 是一款适用于 Apple Silicon 的开源 (Apache 2.0) OpenAI 和 Anthropic 兼容的 LLM 推理服务器,基于 MLX 构建,专注于调用编码代理的可靠工具。在相同重量下,比 Apple 的 MLX (mlx-lm) 快 4 倍。