FastMLX v0.1.0, a high-performance server designed for hosting MLX models on Mac, has been officially unveiled. Developed by Prince Canuma, FastMLX supports both Vision Language Models (VLMs) and Language Models (LMs) and features an OpenAI-compatible API, asynchronous calls, and parallel calls enabled by default. The server allows multi-agent parallelism, concurrent chat handling, and cross-model execution, making it possible to run vision and language models in parallel. Additionally, FastMLX can scale to MacBook specifications, enhancing efficiency. This release positions FastMLX as a potential alternative to other MLX platforms like Ollama and llama.cpp-server, leveraging Apple Silicon’s unified memory for efficient local multi-agent setups. The server operates under the Apache 2.0 License.