MiniMax-M3.1-Flash Gets a 2-Bit GGUF for Local Deployment
Unsloth AI and MiniMax report that Dynamic GGUF quantization enables local deployment of the 428B-parameter MiniMax-M3.1-Flash model. MiniMax says the 2-bit version has a 143.9GB footprint and 91.5% top-1 token agreement across seven domains, while Unsloth reports 91.5% retained top-1 accuracy, a 45% size reduction and operation on a system with 150GB of RAM. The model can be self-hosted through Unsloth Desktop and llama.cpp, with an OpenAI-compatible API available for existing applications and agents.
The release materially expands self-hosting options for MiniMax-M3.1-Flash by reducing its local deployment footprint, although the size and accuracy figures are vendor-reported.
MiniMax-M3.1-Flash can now run locally! We made 2-bit GGUF retain 91.5% top 1 accuracy while being 45% smaller. Run on 150GB RAM. MiniMax-M3.1-Flash is the strongest open model for its size & among the best overall. GGUF: https://huggingface.co/unsloth/MiniMax-M3.1-Flash-GGUF Guide: https://unsloth.ai/docs/models/minimax-m3.1-flash
MiniMax (official): MiniMax M3.1 Flash: Space Performance + Bunny Speed 🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash. 💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence
#MiniMax#Unsloth AI#MiniMax-M3.1-Flash#Unsloth Desktop#llama.cpp