Model tracker / Cohere
North Mini Code
Open-weight 30B MoE model (3B active) for agentic coding and terminal tasks
North Mini Code is a fast open-weight model from Cohere, released 9 June 2026. It's free at Cohere, with a 256K-token context window.
- Made by
- Cohere
- Released
- 2026-06-09
- Status
- Generally available
- Tier
- Fast
- Weights
- Open
- Memory to run
- About 74 GB (BF16)
- Context window
- 256,000 tokens
- Max output
- 64,000 tokens
- Input
- text
- API id
north-mini-code-1.0
North Mini Code price at Cohere
US dollars per million tokens. Blended is 3 input tokens for every output token, one number to compare on.
| Provider | Input | Output | Cached input | Blended | Batch | Checked |
|---|---|---|---|---|---|---|
| Cohere · maker | Free | Free | – | Free | – | 2026-11-02 |
- Cohere: Cohere lists North Mini Code as Free: trial and production keys are free until rate limits are reached. Cohere describes it as an open-weights research release, offered on its API for evaluation; it can also be deployed in production through Cohere's Model Vault (billed per instance; no North Mini Code rate listed). No cached-input or batch price.
North Mini Code benchmarks
| Benchmark | Score | What it measures | Reported by |
|---|---|---|---|
| SWE-bench Verified | 67.6% | real GitHub issues fixed | Cohere |
Scores reported by a model's maker aren't independently checked.
Can you run North Mini Code yourself?
Yes. Cohere publishes the weights on Hugging Face as CohereLabs/North-Mini-Code-1.0, under Apache 2.0: 30B (MoE, 3B active) parameters, 61 GB in BF16.
Running it takes about 74 GB of GPU memory: one 80 GB data-centre GPU, or a Mac with comfortably more unified memory than that. That's the weight files plus 20% for working memory with short inputs; long contexts need more, and fitting in memory doesn't make it fast.
Cohere also publishes a smaller FP8 version, CohereLabs/North-Mini-Code-1.0-fp8, at 32 GB: about 39 GB to run, so one 80 GB data-centre GPU, or a Mac with comfortably more unified memory than that.
Models like North Mini Code
Other fast models closest in price, at each maker's own price or the cheapest provider.
| Model | Input | Output | Context | Weights |
|---|---|---|---|---|
| North Mini Code | Free | Free | 256K | Open |
| gpt-oss-20b OpenAI | $0.03 | $0.14 | 131K | Open |
| gpt-oss-120b OpenAI | $0.037 | $0.17 | 131K | Open |
| NVIDIA Nemotron 3.5 Lightning 30B A3B NVIDIA | $0.06 | $0.16 | 262K | Open |
| Ministral 3 3B Mistral AI | $0.10 | $0.10 | 256K | Open |
| LFM3-1.7B Liquid AI | $0.08 | $0.20 | 262K | Open |
Compare them all in the model tracker →
North Mini Code FAQ
How much does North Mini Code cost?
North Mini Code is listed as free at Cohere. Prices are list prices in US dollars, last checked 2 November 2026.
What is North Mini Code's context window?
North Mini Code takes up to 256,000 tokens in one request and writes up to 64,000 tokens in one response.
Is North Mini Code open weight?
Yes. Cohere publishes the weights, so you can run North Mini Code yourself or at any host that serves it.
Can I run North Mini Code locally?
North Mini Code's weights are 61 GB in BF16, so it needs about 74 GB of GPU memory: one 80 GB data-centre GPU, or a Mac with comfortably more unified memory than that. Cohere also publishes a FP8 version that needs about 39 GB.
Is North Mini Code free?
At Cohere, trial and production keys are free up to their rate limits.
Where can I use North Mini Code?
North Mini Code is available from Cohere. At Cohere, call it as north-mini-code-1.0.