Mistral logo

Devstral 2 123B

Devstral 2 123B loads in 74.9 GB at Q4_K_M, comfortably from 128 GB of memory, Macs included, starting at the Mac Studio M5 Ultra 256GB. Below is the memory math we ran for it, the cheapest card that fits, and the fastest machine we track.

PARAMETERS
123B
FORMAT
Q4_K_M
MIN MEMORY
96 GB
BEST FOR
Agentic coding, Software engineering

How much memory does Devstral 2 123B need?

Devstral 2 123B is a 123B dense model built for agentic coding, software engineering. The weights alone take 74.9 GB at Q4_K_M; add context and the OS, and it belongs in the 128 GB memory class. Every number below comes from the ModelFit engine — speeds are estimates, labeled est., because we do not benchmark hardware ourselves.

Weights (Q4_K_M)
74.9 GB
+ KV cache (16k)
~6.0 GB
Total at 16k
~80.9 GB
Comfortable from
128 GB

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

Which machines run Devstral 2 123B?

Cheapest GPU
~14 tok/s est. — ~$12,912 used (as of 2026-07-31)
Fastest
~14 tok/s est.
Runs on a Mac?
~8 tok/s est.

Verdicts come from the same engine that powers our fit checker — memory capacity first, bandwidth for speed. How we compute these numbers · Hugging Face model card

More Mistral models

Frequently asked questions

How much memory does Devstral 2 123B need?

74.9 GB for the Q4_K_M weights, plus ~6.0 GB of KV-cache at 16k context, about 80.9 GB total. Comfortable from 128 GB of VRAM or unified memory.

Does Devstral 2 123B run on a Mac?

Yes, from the Mac Studio M5 Ultra 256GB (~8 tok/s est.). Unified memory means the RAM budget is the only limit.

What is the cheapest GPU for Devstral 2 123B?

The NVIDIA RTX PRO 6000 Blackwell is the cheapest tracked card that runs Devstral 2 123B comfortably, ~14 tok/s est. at ~$12,912 used (as of 2026-07-31).

Cite this page

ModelFit: Devstral 2 123B, specs, memory math and hardware verdicts.
https://modelfit.io/models/devstral-2-123b/ (dataset updated 2026-09-29, CC BY 4.0).