Key Info

Unsloth has released Dynamic GGUFs for MiniMax M3.1 Flash, bringing the 428B MoE model down to a 143.9GB 2-bit footprint while retaining 91.5% top-1 token agreement in Unsloth's evaluation across seven domains.

Highlights

  • The 2-bit GGUF is roughly 45% smaller and runs on about 150GB of RAM, lowering the memory barrier to self-hosting.
  • Deployment is supported through Unsloth Desktop and llama.cpp, with an OpenAI-compatible API for existing apps and agents.
  • Unsloth cites 78K+ GitHub stars and over 500M Hugging Face model downloads as the community reach for this release.
  • MiniMax frames the release as extending open-weight, local-control options for developers.