Mellum2.1, a fast open model for coding agents
JetBrains has released Mellum2.1, the next version of the 12B mixture-of-experts model it open-sourced in June. The architecture is unchanged, with 2.5B active parameters under Apache 2.0; the new part is a summer of reinforcement learning in real environments, aimed at coding agents and fast sub-agents on your own hardware. The weights are on Hugging Face, and GGUF builds are, the maker says, coming soon.
Key facts
- Model
- Mellum2.1
- Architecture
- 12B mixture-of-experts, 2.5B active parameters
- Licence
- Apache 2.0
- Built for
- Coding agents and fast sub-agents on your own hardware
- Training
- Reinforcement learning in real environments
- Where
- Hugging Face
- Coming soon
- GGUF builds for llama.cpp, Ollama and LM Studio
- Listed
- 8 October 2026
- Source
- JetBrains
Our note above; facts as published by JetBrains.