Cohere Releases North Mini Code, Open-Source Model for Agentic Software Engineering

Summary

Cohere announced North Mini Code, a 30-billion-parameter Mixture-of-Experts model with 3 billion active parameters designed specifically for agentic software engineering and complex code generation. The model is released on Hugging Face under the Apache 2.0 license, making it freely available to developers worldwide.

North Mini Code achieves impressive performance on coding benchmarks, scoring 33.4 on Artificial Analysis' Coding Index and outperforming several much larger open-source models including Qwen 3.5 (35B active parameters), Mistral Small 4 (119B active parameters), and Devstral 2 (123B active parameters). The model is optimized for terminal-based agentic tasks and complex software engineering workflows.

Cohere trained North Mini Code using a two-stage cascaded supervised fine-tuning approach followed by reinforcement learning with verifiable rewards (RLVR), ensuring the model works reliably across different agent harnesses. The training leveraged over 70,000 verifiable tasks from approximately 5,000 unique repositories, with special emphasis on making the model robust for real-world coding agents.

• Novel training approach combines supervised fine-tuning with reinforcement learning using verifiable rewards from 70k+ real-world coding tasks
• Designed to work reliably across multiple agent harnesses and scaffolds for production-grade coding agents

Editorial Opinion

North Mini Code represents a meaningful shift toward making agentic AI coding practical and accessible. By open-sourcing a specialized model that outperforms significantly larger alternatives, Cohere is lowering the barrier for developers to build production-grade AI agents. The emphasis on verifiable rewards and training on real-world repositories signals a mature approach to building agents that actually work in practice, rather than just chasing benchmark scores.