Cohere North Mini Code

Cohere North Mini Code loads in 19 GB at Q4_K_M and is comfortable from 36 GB of memory, Macs included from the MacBook Pro M5 Pro 48GB. Below: the full memory math, the cheapest card that runs it well, and the fastest machine we track.

PARAMETERS
30B (3B active)
FORMAT
Q4_K_M
MIN MEMORY
32 GB
BEST FOR
Agentic coding, repo-scale software engineering

Memory math

Weights (Q4_K_M)
19 GB
+ KV cache (16k)
~4.0 GB
Total at 16k
~23.0 GB
Comfortable from
36 GB

KV = fp16 estimate (q8_0 cache roughly halves it). "Comfortable" = weights + KV within the engine's tiered budget (~70-85% of memory).

Hardware snapshot

Cheapest GPU
~77 tok/s est. — ~$700 used
Fastest
~125 tok/s est.
Runs on a Mac?
~42 tok/s est.

Frequently asked questions

How much memory does Cohere North Mini Code need?

19 GB for the Q4_K_M weights, plus ~4.0 GB of KV-cache at 16k context — about 23.0 GB total. Comfortable from 36 GB of VRAM or unified memory.

Does Cohere North Mini Code run on a Mac?

Yes — from the MacBook Pro M5 Pro 48GB (~42 tok/s est.). Unified memory means the RAM budget is the only limit.

What is the cheapest GPU for Cohere North Mini Code?

The AMD Radeon RX 7900 XTX is the cheapest tracked card that runs Cohere North Mini Code comfortably — ~77 tok/s est. at ~$700 used.

Cite this page

ModelFit: Cohere North Mini Code — specs, memory math and hardware verdicts.
https://modelfit.io/models/north-mini-code-1/ (dataset updated 2026-09-03, CC BY 4.0).