Cohere · AI model
North Mini Code
Cohere's first agentic coding model (June 2026), a 30B total / 3B active parameter mixture-of-experts model; 256K context, 64K output; text input; Apache 2.0 weights; free on the Cohere API up to the rate limits.
Last verified Available 4 sources
Key specs
- Provider
- Cohere
- Family
- North
- API identifier
- north-mini-code-1-0
- Availability
- Available
- Release date
- 2026-06-09
- License
- Open source (Apache 2.0)
- Context window
- 256,000 tokens
- Max output
- 64,000 tokens
- API access
- Yes
- Consumer app
- No
Capabilities
Inputs
- Text input Yes
- Image input No Image Inputs marked as not supported in the capabilities panel on the model's docs page; model card: input text only
- Audio input No Model card: input text only
- Video input No Model card: input text only
- File input Unknown
Outputs
- Text output Yes
- Image output No Model card: model generates text
- Audio output No Model card: model generates text
Features
- Reasoning Optional Reasoning listed in the capabilities panel on the model's docs page; model card: supports interleaved thinking and 'works best when turned on'; Cohere's Reasoning guide states its reasoning models are hybrid. An API-level toggle is not documented specifically for this model
- Tool calling Yes Tool Use listed in the capabilities panel on the model's docs page; documented for terminal-based agents and harnesses such as SWE-Agent and OpenCode
- Web search Unknown
"Unknown" means we have not verified this capability yet; it does not mean the model lacks it.
Pricing
No current prices listed.
Benchmarks
| Benchmark | Setting | Score | Reported by | Measured | Verified | Source |
|---|---|---|---|---|---|---|
| GDPval-AA v2.1 Elo, higher is better | GDPval-AA v2.1; effort not stated (row label 'North Mini Code'); Stirrup agent harness with shell and web browsing | 311 Elo | Third party | 2026-10-25 | 2026-10-25 | GDPval-AA v2.1 Leaderboard (Artificial Analysis) |
Elo is a relative rating: values change as models are added and the scale is re-anchored, so always store the date a rating was read. For example, Anthropic's launch page gives 1846 for Claude Opus 5.5, while the leaderboard on 2026-10-12 shows 1866 (Max, Default Fallback). Earlier versions (Google cites "GDPVal-AA v2") are not comparable. Judging is done by an LLM, not by human experts. Anthropic models are listed with "Default Fallback", meaning refused steps are routed to a fallback model. AA-Briefcase and GDP.pdf are separate evaluations.
One row per result, without rank numbers: each result was measured under its own setting (effort level, harness, tools, test version), as stated by the source, and scores compare only within the same setting. Self-reported results come from the provider and have not been reproduced independently.
Compare North Mini Code
Guides
About North Mini Code
North Mini Code is a 30B total / 3B active parameter mixture-of-experts model trained for agentic coding, released on June 9, 2026 as the first model in Cohere's North family of code agent models. It targets repository-level code changes in harnesses such as SWE-Agent and OpenCode, terminal-based agents and local deployment. Weights are released under Apache 2.0; on the Cohere API (Chat V2) it is offered for evaluation and is free for trial and production keys up to the rate limits, with production deployment via Model Vault.
Strengths and weaknesses
Strengths
- Trained for agentic software engineering and terminal-based agents; supports tool use, structured outputs and reasoning
- 256K-token context window and up to 64K output tokens
- 3B active parameters, aimed at low-latency local inference
- Open weights under Apache 2.0
Weaknesses
- Specialized for coding tasks; text-only input
- Available only on the Chat V2 endpoint (not Chat V1 or the Chat Completions compatibility endpoint)
- No paid per-token API price is published; Cohere API access is described as being for evaluation
Sources
Last verified Page updated 2026-10-26