Cohere releases North Mini Code for coding agents
Cohere released North Mini Code 1.0. It is a 30B open-weights model built for coding agents and terminal tasks. The model is available under the Apache 2.0 license on Hugging Face, and it can be used through OpenCode, the Cohere API, and Hugging Face. Community posts also discussed benchmark scores, local use, GGUF builds, and the current llama.cpp support status.
Key points
- Cohere released North Mini Code 1.0 for coding agents.
- The model has 30B total parameters and 3B active parameters.
- It is released as open weights under the Apache 2.0 license.
- It is available through OpenCode, the Cohere API, and Hugging Face.
- GGUF builds may help local use, but llama.cpp support needs checking before adoption.
Quick term guide
- open-weight
- The model's internal numbers are publicly released, so anyone can download and run or modify it freely.
- coding agents
- AI programs designed to autonomously perform tasks like writing or fixing code.
- coding agent
- An AI tool that writes or edits code from a person’s instructions.
- Hugging Face
- An online place where AI models and datasets are shared.
- 3B active parameters
- Only about 3 billion parts of the model are used for each step, even though the full model is larger.
- active parameters
- The part of the model that is actually used for a given answer.
- pull request
- A formal way to propose code changes and ask others (or an AI) to review them before they're merged into the main codebase
- open weights
- Model files are shared so others can download and run the AI model themselves.