Cohere Open-Sources Coding Agent for Single H100

Cohere has launched North Mini Code, an open-source 30 billion parameter mixture-of-experts model designed for agentic coding tasks. It runs on a single H100 GPU and supports a 256,000 token context window, making it a self-hostable alternative to managed models like Claude. It is available on Hugging Face under an Apache 2.0 license. The model handles sub-agent orchestration, code review, architecture mapping, and terminal work. However, independent testing found it generates three times more output tokens than comparable models, a verbosity drawback that could raise costs significantly in high-volume production environments.