Cohere open-sources North Mini Code agent, runs on single H100, ranks 8th in speed
The AMW Read
Cohere enters open-weight coding agent space with a single-GPU optimized model, updating the player map for AI coding tools segment with a new entrant that competes with Cursor and Windsurf.
Cohere open-sources North Mini Code agent, runs on single H100, ranks 8th in speed
Cohere has released an open-weight coding agent called North Mini Code, designed to run on a single NVIDIA H100 GPU. The agent ranks 8th among open-weight models for output speed and produces 3x more tokens compared to typical coding agents, according to the company.
This release marks Cohere's entry into the open-weight coding agent space, a segment currently dominated by specialized players like Cursor and Windsurf. By open-sourcing North Mini Code and optimizing it for single-GPU deployment, Cohere is lowering the barrier to entry for developers and enterprise teams who want to deploy a capable coding agent without massive compute infrastructure.
The move aligns with a recurring pattern in the AI coding tools segment: the acqui-licensing and open-weight distribution play, where model labs use open releases to drive ecosystem adoption and enterprise procurement pipelines. Cohere's decision to open-source a competitive coding agent that runs on commodity hardware could pressure incumbents to either lower pricing or improve inference efficiency. It also signals that Cohere sees the coding agent market as a strategic distribution channel for its broader enterprise platform, rather than a standalone product.
