North Mini Code is an AI Code tool. Open-source coding model for agentic software engineering and code generation tasks. Key features include Mixture-of-Experts Architecture, Agentic Software Engineering, and Large Context Window. Best for software developers and engineers and data scientists and analysts.
About North Mini Code
Key Features
<strong>Mixture-of-Experts Architecture.</strong> Uses 30B total parameters with only 3B active at any time, reducing hardware requirements while maintaining strong performance for code generation and software engineering tasks.
<strong>Agentic Software Engineering.</strong> Built specifically for agentic workflows including sub-agent orchestration, systems architecture mapping, and repo-level code changes across multi-turn tasks.
<strong>Large Context Window.</strong> Supports 256K token context window with 64K token output, allowing the model to process entire mid-sized codebases and generate substantial code blocks in a single pass.
<strong>Open-Source Apache 2.0 License.</strong> Freely available on Hugging Face with full model weights, enabling developers to run the model locally or on-premise without sending data to external clouds.
<strong>Terminal-Based Agent Capabilities.</strong> Designed to drive shell tools end-to-end across multi-turn tasks, with strong performance on terminal benchmarks and command-line workflows.
<strong>Fast Inference Speed.</strong> Delivers approximately 199 output tokens per second on Cohere's API, with up to 2.8x higher throughput than comparable models while maintaining competitive latency.
Frequently Asked Questions
North Mini Code is designed for agentic software engineering tasks including code generation, repo-level code changes, architecture mapping, code review, and terminal-based workflows. It works with coding agents like OpenCode and can handle complex multi-turn programming tasks.
Yes, North Mini Code is completely free and open-source under the Apache 2.0 license. You can download the model weights from Hugging Face and run it locally, or use it through Cohere's API for free until rate limits are reached.
North Mini Code has 30B total parameters but only 3B active at inference time, making it efficient enough to run on a single H100 GPU. The Mixture-of-Experts architecture reduces computational overhead compared to dense models of similar size.
North Mini Code scores 33.4 on the Artificial Analysis Coding Index, performing competitively against similar-sized open-source models. It outperforms models like Devstral Small 2 and even some larger models, though it generates more output tokens than comparable models.
