Part of 786+ curated AI tools on AISO
North Mini Code logo

North Mini Code

Cohere's open-source agentic coding model — 30B MoE, 3B active, Apache 2.0

0
freemiumDR 81Apache 2.0 open weights — free to download and self-host from Hugging Face. Available via Cohere API (pay-per-token), Cohere Model Vault (dedicated managed inference), and OpenRouter. Minimum hardware: 1× H100 @ FP8.View full pricing →

Visit North Mini Code

https://cohere.com/blog/north-mini-code

About North Mini Code

Cohere's first agentic coding model and inaugural member of the North model family. A 30B Mixture of Experts model with only 3B active parameters per token, released June 9, 2026 under Apache 2.0. Achieves 2.8× higher output throughput than Devstral Small 2 on identical hardware, 256K context, and runs on a single H100 at FP8.

Key Features

30B total / 3B active MoE architecture — dense-model quality at fraction of inference cost
2.8× higher output throughput than Devstral Small 2 (identical hardware)
30% better inter-token latency than Devstral Small 2
33.4 on Artificial Analysis Coding Index
256K total context window; 64K max generation
Apache 2.0 license — fully open for commercial use, modification, and redistribution
Single H100 @ FP8 minimum — unusually accessible for a 30B model
Optimized for code generation, agentic software engineering, and terminal tasks
Available on Hugging Face, Cohere API, Model Vault, OpenRouter, and OpenCode

North Mini Code Pros & Cons

Pros

  • +Apache 2.0 license — genuinely free for commercial use with no additional restrictions
  • +Single H100 minimum requirement makes self-hosting accessible for small teams and solo developers
  • +2.8× throughput advantage over Devstral Small 2 means dramatically lower cost-per-coding-task at scale
  • +256K context handles large codebases and extended agentic sessions without chunking
  • +MoE architecture (3B active) keeps inference costs comparable to a 3B dense model
  • +Cohere's enterprise-grade infrastructure available via API for teams that don't want to self-host

⚠️ Cons

  • Newer model with limited community tooling and production track record compared to Llama or Mistral
  • Focused on coding only — not a general-purpose model for chat, reasoning, or multimodal tasks
  • Cohere has historically been enterprise-first; developer ecosystem support is still maturing
  • No published scores on SWE-Bench Verified — harder to compare directly to Mistral Medium 3.5 or Claude
  • Time-to-first-token is slightly behind Devstral Small 2 per Cohere's own testing

Who Is North Mini Code Best For?

👤Developers who want a high-throughput open-source coding model on a single GPU
👤Teams building agentic software engineering pipelines that need long-context code generation
👤Organizations in regulated industries needing sovereign, on-premise coding AI
👤Cost-sensitive workloads where MoE inference efficiency translates directly to lower GPU bills

Tags

coherellmcodingopen-weightsmixture-of-expertsagenticself-hostedapache 2.0developer tools
🏷️

Is this your tool?

Claim your listing to get a Featured badge, edit your description, and stand out from competitors. All plans include a permanent dofollow backlink to your site.

Claim Now →

📬 Get the best new AI tools delivered weekly

One concise email with fresh launches, trending picks, and featured standouts.

Alternatives to North Mini Code

View all North Mini Code alternatives →