Millie 35B-A3B 9GB

Millie is a 35B-parameter mixture-of-experts coding and agentic model (3B active per token, 262K context, image input) compressed to run on ordinary hardware. It is derived from Agents-A1, built on the Qwen3.5-35B-A3B architecture; expert weights are stored in a ternary mask-and-sign format at 1.69 bits per weight, making this the smallest and fastest Millie variant. Larger 11 GB and 7 GB variants are also available.

The GGUF files require the llmsforall llama.cpp fork, which carries the kernels for their weight format (Metal, Vulkan, CPU); upstream llama.cpp cannot load them:

llama-server -m Millie-35B-A3B-9GB.gguf --mmproj Millie-35B-A3B-mmproj.gguf

License

Apache 2.0 — see LICENSE and NOTICE.