Run the right open model on the right compute.
Drop in our OpenAI-compatible client and specify where each inference runs, in the cloud or on-device.
Tap into the open model ecosystem
Run models from Hugging Face and GitHub through one standardized interface. Muna handles model compilation and deployment, so you can iterate quickly and deploy confidently.
With inference you can place anywhere
Our compiler converts AI models into portable artifacts optimized for every target. You decide, in code, where your inference runs: from a remote_b200to your local_gpu.
OpenAI experience, self-hosted economics
We provide client libraries for all major frameworks and platforms. JavaScript, Python, Swift, Kotlin, Flutter, and more.
