North Mini Code is Cohere's June 2026 code-specialist model from the North line, and inside Atlas it is the one Cohere checkpoint that can actually emit a large multi file refactor in a single turn. North Mini Code carries a 256,000 token context with a 64,000 token output, the largest generation budget Cohere ships, and it is reasoning enabled. The registry lists it at $0 per Mtok input and $0 per Mtok output, which almost certainly means access rides on a North plan rather than metered tokens, so confirm your entitlement before you build a workflow on it.
What makes North Mini Code different from Command A in Atlas?
North Mini Code's 64,000 token output ceiling is 8x Command A's 8,000, and that single number changes how you use Atlas. Both models carry a 256,000 token context, but only North Mini Code can emit a large multi file refactor in one turn. Cohere shipped it in June 2026.
Every other Cohere checkpoint forces a file-by-file cadence because the generation budget runs out. North Mini Code does not. With 64,000 tokens of output, the largest generation budget Cohere ships, Atlas can take an approved plan and produce the whole change in a single build pass. Atlas's documented guidance leans into exactly that: take advantage of the 64,000 token output by asking Atlas for whole-package refactors in a single plan then build pass. North Mini Code is also reasoning enabled and code-specialized, which is the exact profile Atlas's build agent wants, so unlike the Command family it is not depending on the plan agent to supply thinking it lacks.
Why does the $0 per Mtok listing need a second look?
North Mini Code is listed at $0 per Mtok on both input and output in the registry, making it the cheapest reasoning model in Cohere's lineup to experiment with. But a $0 listing means access is likely gated behind a North plan rather than metered per token, so confirm entitlement before depending on it.
A $0 per Mtok price is not a discount, it is a signal that the billing model lives somewhere other than the token meter. For a team evaluating North Mini Code in Atlas, that has one practical implication: verify your Cohere account actually has North access before you architect a workflow around a 64,000 token output. Atlas's setup path checks this for you. Export COHERE_API_KEY from an account that has North access, then run `atlas models cohere` and confirm north-mini-code-1-0 resolves for your key. If the id does not resolve, the entitlement is missing, and no amount of atlas.json editing will conjure it. Do that check first, not after you have built a refactor pipeline on top of it.
How do you run a whole-package refactor with North Mini Code?
Run whole-package refactors with North Mini Code as a single plan then build pass. Atlas drafts a plan in a read-only plan agent and asks before switching to a build agent, and North Mini Code's 64,000 token output means the build agent can execute that entire plan without splitting it across turns.
The mechanics are worth spelling out because they differ from every other Cohere model in Atlas. First, let Atlas's read-only plan agent draft the change and approve it. Second, switch to the build agent and let North Mini Code emit the full diff. Third, review it: Atlas computes a unified diff for every file edit and surfaces it for approval before writing, so a 64,000 token generation is still gated file by file at the write boundary. Fourth, and this is the step Atlas's setup guidance calls out specifically, use Atlas edit checkpointing (git patches) so a large generated diff can be rolled back cleanly. A big generation is only safe when a big undo exists, and Atlas snapshots file changes as git patches precisely so it does.
Should you trust a new model with a large refactor?
North Mini Code is a new model with a short track record compared to the Command A family, which shipped in March 2025. Cohere released North Mini Code in June 2026. A 64,000 token output is powerful, and on an unproven checkpoint it is also 64,000 tokens of unreviewed change.
Treat North Mini Code's newness as a process requirement, not a disqualifier. The Command A family has more miles on it. North Mini Code has more capability. The way Atlas reconciles those is through gates rather than trust. Every Atlas tool call is permission-gated against allow, ask, and deny rules before it runs, so you can require approval on writes while you build confidence in the model. Atlas snapshots file changes as git patches so edits can be diffed and rolled back, which makes a bad 64,000 token generation an inconvenience rather than an incident. And Atlas reads git branches, status, and diffs, and can stage and create commits on your behalf, so you can land a verified refactor as a clean commit once you have actually read it.
When should you pick a different model than North Mini Code?
Pick a different model than North Mini Code when entitlement is uncertain or the track record matters more than the output budget. North Mini Code's $0 per Mtok registry listing likely reflects North plan gating, and Command A's 256,000 token context at $2.5 per Mtok is a proven fallback.
Two clear cases. First, if `atlas models cohere` does not resolve north-mini-code-1-0 for your key, you do not have access, and Command A at $2.5 per Mtok input and $10 per Mtok output gives you the same 256,000 token context with a metered, predictable bill. Second, if you are shipping something where a short model track record is itself the risk, the Command A family is older and better understood, and its 8,000 token output ceiling forces the incremental cadence that a cautious team probably wants anyway. Atlas lets you switch the active model and provider on the fly with favorites and recents, so keeping both north-mini-code-1-0 and command-a-03-2025 pinned lets you match the model to the risk of the task.
Setup
- 01Export COHERE_API_KEY from an account that has North access.
- 02Run `atlas models cohere` and confirm north-mini-code-1-0 resolves for your key.
- 03Pin "model": "cohere/north-mini-code-1-0" in atlas.json.
- 04Take advantage of the 64,000 token output by asking Atlas for whole-package refactors in a single plan then build pass.
- 05Use Atlas edit checkpointing (git patches) so a large generated diff can be rolled back cleanly.
Frequently asked questions
- what is North Mini Code's output token limit
- North Mini Code has a 64,000 token output ceiling, 8x Command A's 8,000 and the largest generation budget Cohere ships. Its context window is 256,000 tokens.
- is North Mini Code really free
- North Mini Code is listed at $0 per Mtok input and $0 per Mtok output in the registry, but that likely means access is gated behind a North plan rather than metered per token. Confirm entitlement first.
- how do I set up North Mini Code in Atlas
- Export COHERE_API_KEY from an account that has North access, run `atlas models cohere` and confirm north-mini-code-1-0 resolves, then pin "model": "cohere/north-mini-code-1-0" in atlas.json.
- can North Mini Code do a multi file refactor in one turn
- Yes. North Mini Code's 64,000 token output means it can emit a large multi file refactor in a single turn. Ask Atlas for whole-package refactors in one plan then build pass.
- does North Mini Code support reasoning
- Yes. North Mini Code is reasoning enabled and code-specialized, which is the exact profile Atlas's build agent wants, unlike the non-reasoning Command family.
- North Mini Code vs Command A for coding
- Both carry a 256,000 token context, but North Mini Code emits 64,000 output tokens against Command A's 8,000, and North Mini Code is reasoning enabled while Command A is not.
- how do I roll back a large AI generated refactor
- Atlas snapshots file changes as git patches so edits can be diffed and rolled back. With North Mini Code's 64,000 token output, use that edit checkpointing on every large generated diff.
Try Atlas in your terminal
The terminal-native AI coding agent. Free core, single binary.
Install AtlasRelated guides
Atlas for Quarkus in 2026
Atlas is a terminal-native AI coding agent for Quarkus in 2026. It reads CDI beans and JAX-RS resources, then runs ./mvnw test behind a permission prompt.
Write Unit Tests for Untested Code with Atlas in 2026
How to write unit tests for untested code with Atlas in 2026: the lsp tool enumerates exported symbols, grep copies repo conventions, and bash actually runs the suite.
Atlas vs Mistral Vibe for Code: Terminal AI Coding Agents in 2026
Compare Atlas and Mistral Vibe for Code in 2026. Atlas offers terminal-native TUI, explicit diffs, and BYO models. Mistral Vibe for Code provides a four-model stack, multi-platform access, and EU data sovereignty.
Plan a Multi-File Change Before Editing with Atlas in 2026
How to plan a multi-file change with Atlas in 2026: the plan agent denies all edit tools, you research with codebase_search and lsp, then plan_exit hands off.
Atlas for Scala in 2026
Atlas is a terminal-native AI coding agent for Scala in 2026. Run it in a project with a build.sbt, let it read your traits and implicits, and approve every diff.
Atlas vs Cline: Terminal AI Coding Agents in 2026
Compare Atlas and Cline, two leading AI coding agents for 2026. Atlas offers terminal-native TUI and permission-gated tools, while Cline integrates with VS Code and provides checkpoint rollbacks.
Atlas for Fastify in 2026
Atlas is a terminal-native AI coding agent for Fastify in 2026. It reads plugin encapsulation and JSON schemas, then runs node --test behind a permission prompt.
Atlas vs Gemini CLI: A Developer's Guide to Terminal AI Agents in 2026
Atlas vs Gemini CLI in 2026: Compare terminal AI coding agents. Atlas offers permission-gated tool calls and diff-based approvals. Gemini CLI provides a million-plus token context and a free tier.