GPT-OSS 120B

GPT-OSS 120B loads in 65.4 GB at MXFP4 and is comfortable from 96 GB of memory, Macs included from the MacBook Pro M5 Max 128GB. Below: the full memory math, the cheapest card that runs it well, and the fastest machine we track.

PARAMETERS
117B (5.1B active)
FORMAT
MXFP4
MIN MEMORY
96 GB
BEST FOR
Reasoning, Coding, Agents

Memory math

Weights (MXFP4)
65.4 GB
+ KV cache (16k)
~6.0 GB
Total at 16k
~71.4 GB
Comfortable from
96 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
~11 tok/s est. — ~$3,847 used (as of 2026-07-31)
Fastest
~51 tok/s est.
Runs on a Mac?
~29 tok/s est.

Go deeper

Every GPT-OSS 120B quant by real GGUF file sizeBest hardware for GPT-OSS 120B: cheapest, best value, fastest

More GPT-OSS models

GPT-OSS 20B

Frequently asked questions

How much memory does GPT-OSS 120B need?

65.4 GB for the MXFP4 weights, plus ~6.0 GB of KV-cache at 16k context — about 71.4 GB total. Comfortable from 96 GB of VRAM or unified memory.

Does GPT-OSS 120B run on a Mac?

Yes — from the MacBook Pro M5 Max 128GB (~29 tok/s est.). Unified memory means the RAM budget is the only limit.

What is the cheapest GPU for GPT-OSS 120B?

The AMD Ryzen AI Max+ 395 (Strix Halo) is the cheapest tracked card that runs GPT-OSS 120B comfortably — ~11 tok/s est. at ~$3,847 used (as of 2026-07-31).

Cite this page

ModelFit: GPT-OSS 120B — specs, memory math and hardware verdicts.
https://modelfit.io/models/gpt-oss-120b/ (dataset updated 2026-09-03, CC BY 4.0).