Cohere releases North Mini Code, its first open-source coding model under the Apache 2.0 license
The compact model targets agentic software development workflows
Many users praised Cohere's open-source North Mini Code model for its Apache 2.0 license enabling local deployment in sensitive fields like healthcare, while a few questioned its performance claims and marketing.
Most Activity
This is just awesomeness from @cohere, @nickfrosst, and team.
I so badly want a coding agent that just runs on my local machine.
We are not too far now!
Excited to get this to work with my @dair_ai coding agent in the next couple of days.
Introducing Cohere's first open-source coding model: North Mini Code
Small & efficient, designed for agentic performance and built for community input.
this model is the opposite of mythos.
Its small, cost effective, apache 2.0, and locally deployable. This is the way LLMs should go.
small, open source, transparent and sovereign vs large, expensive, proprietary and hegemonic
Introducing Cohere's first open-source coding model: North Mini Code
Small & efficient, designed for agentic performance and built for community input.
The Apache 2.0 + local deployment combination is exactly what healthcare AI needs right now. Clinical notes can't go to a cloud API in most production deployments. A small, efficient, locally runnable coding model changes what's actually buildable in regulated environments. Congrats on the launch @nickfrosst @cohere. The "built in public" framing resonates.
honoured to work with some of the best hair in the office for this model release
Introducing Cohere's first open-source coding model: North Mini Code
Small & efficient, designed for agentic performance and built for community input.
@nickfrosst Tried it in opencode, very sweet. Super fast!
@nickfrosst Security, governance, policy, configurability all belong in the hands of users, not providers.
@nickfrosst This is lovely, thank you!
@nickfrosst Small, local, and auditable still wins for anything that touches sensitive code or needs reproducible behavior. The frontier models are for exploration, not the final pipeline.
@nickfrosst Agreed nick! Awesome work by you and your team congrats on the launch! Looking forward to more models from Cohere. Godspeed to you guys 💯
@nickfrosst Interesting model, but how does it stack up to the competition? In your demo video the model failed half the tool calls lol
@nickfrosst Question for you: are you planning to develop a harness for it ? Cohere Code 💪 🇨🇦 - or is that on the roadmap?
That sounds like a fun project.
@nickfrosst Looking great Nick testing it locally great tool use!
@nickfrosst agree with the framing but the "sovereign" part only works if someone actually runs it
still, small models have the most interesting roadmap right now
@nickfrosst you said it first tbh
question is if the open models catch up on reliability before big labs lock in the user base
@nickfrosst People will use what is most useful for them based on cost/speed/accuracy trade off. Cheap, fast, mediocre open models will always have a place in the harness.
@nickfrosst Is "built for community input" product marketese for "this shit doesn't work yet"?
@nickfrosst How does it compare to the other open source models?
@nickfrosst Totally agree. However, I’m happy for the masses to fund Anthropic/OpenAI whilst the real developers take advantage of these open LLMs. After a few days of bedding in, I’ve got OpenCode producing some truly excellent code from the open models, albeit hosted vs local.
@nickfrosst TRVE, but man… fable is awe-inspiring.
@nickfrosst Cohere flips the narrative on Mythos...
Cohere Atom / Atoma: The indivisible, fundamental building block of matter. It highlights that the model is small, sovereign, and an essential unit for local deployment.