Argonaut Labs · Local inference research

Large models.
Local hardware.

We build open-source inference engines and storage tools for running large mixture-of-experts models from NVMe.

Argodrive DS4 Kimi K3.1 includes the Kimi port, expert caches and storage scheduler. Its installation guide covers weights, hardware requirements and setup; the separate Argodrive dashboard is optional.

Read the measured results and hardware conditions, explore our research, or visit our Hugging Face organization profile.