Developers wanting a fast open coding model that fits on one GPU can try it free; Cohere's own API price isn't stated.
Kenning's take
What it does
North Mini Code is Cohere's first agentic coding model and first open-source model for developers. It has 30B total and 3B active parameters, a 256K context window and runs on a single H100 GPU, and Cohere says it generates output up to 2.8 times faster than Devstral Small 2. The weights are free under Apache 2.0 on Hugging Face, and it is also available through the Cohere API, Model Vault, OpenRouter and OpenCode.
- 30B total and 3B active parameters
- 256K context, up to 64K generated tokens
- Apache 2.0 weights in bf16, fp8 and w4a16
- Runs on one H100 GPU
- Built for code generation and terminal tasks
Who it suits. Developers who want a small open coding model they can run on their own hardware.
Key facts
- Category
- AI models
- Made by
- Cohere
- Released
- 9 June 2026
- How to get it
- open weights on Hugging Face; through the Cohere API, Cohere Model Vault, OpenRouter and OpenCode
- Works on
- API, self-hosted
- In the UK
- Yes
- Free plan
- free open weights
FAQ
Is North Mini Code open source?
Yes. Cohere released it under the Apache 2.0 licence with weights on Hugging Face.
When was North Mini Code released?
Cohere launched it on 9 June 2026.
What hardware does North Mini Code need?
Cohere lists a minimum of one H100 GPU at FP8 or FP4.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.