LocalClaw/LLM/North Mini Code 1.0
Open-weight MoE

North Mini Code 1.0

Cohere Labs Apache 2.0 coding and agent model. 30B total / 3B active MoE, 256K context, terminal-task training and mature GGUF quantizations for local workstation use.

32 GB power user 32 GB RAM Q4_K_M Local coding assistant
Run with LocalClaw Compare all models
Parameters
30B (3B active, MoE)
Minimum RAM
32 GB
Model size
18 GB
Quantization
Q4_K_M

Can North Mini Code 1.0 run locally?

North Mini Code 1.0 belongs on 32 GB machines when you want stronger quality without jumping to server hardware.

Search for north-mini-code-1.0 in LM Studio or another GGUF-compatible runtime.

Model sourceunsloth/North-Mini-Code-1.0-GGUF
codeagentreasoningpower

Install path

01
Check RAM fitMinimum 32 GB RAM. Start with the Q4_K_M quant.
02
Load the modelSearch north-mini-code-1.0 in LM Studio.
03
Control locallyUse LocalClaw to manage models, agents, chat, channels and scheduled OpenClaw work.

Strengths

  • Official Cohere Labs Apache 2.0 open-weight release
  • Strong fit for coding agents, terminal tasks and SWE-style workflows
  • Sparse 30B-A3B shape keeps active compute close to smaller local models
  • 256K context window and very long output budget for repository-scale work
  • High-signal GGUF availability through Unsloth and Bartowski for LM Studio and llama.cpp style local use
  • Useful local alternative to heavier proprietary coding-agent models

Limitations

  • Specialized for code and agentic tasks rather than being the safest general chat default
  • Full 256K context can require far more memory than the base Q4 footprint suggests
  • Best tool-use behavior may depend on recent runtimes and correct chat-template handling
  • 30B total weights still put it in the 32GB+ workstation tier for comfortable local use

Best use cases

  • Local coding assistant
  • Terminal and shell-task agents
  • Repository-scale code review and debugging
  • SWE-bench style experiments
  • Tool-calling workflows with private code
  • Long-context codebase navigation

Capability profile

speed
5
quality
8
coding
9
reasoning
8

Technical notes

Developer
Cohere Labs
License
Apache 2.0
Context window
262,144 tokens
Architecture
Decoder-only sparse Mixture-of-Experts model with 30B total parameters, about 3B active parameters, 128 experts and 8 activated experts per token.

This model fits these next steps

Hardware fit is based on LocalClaw's RAM tier, model size and quantization metadata. Always leave memory headroom for your OS and runtime.

Comfortable headroomMac mini M4 Pro 48GBMobile workstationMacBook Pro M4 Max 36GBPower-user picks32GB RAM guide

Similar models to compare

Qwen AgentWorld 35B-A3B 35B (3B active, MoE)Qwen 3.6 35B-A3B 35B (3B active, MoE)OLMo 3 32B Think 32BSarvam 30B 32B (2.4B active, MoE)

Where to go next