Only with CPU offload

gpt-oss 120B at MXFP4 needs 61.4 GB but only 18.4 GB is addressable, so about 71% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 7.0 tokens per second — usable for batch work, painful for chat.

29% on GPU 8K context MXFP4 · 60.5 GB Apache 2.0 Released Aug 2025 Native MXFP4

Designed to land on one 80 GB card. Only ~5B parameters are active per token.

Hardware requirements What hardware do I need to run gpt-oss 120B? →

At MXFP4 with an 8K context: a 80 GB GPU (H100 SXM) or a 128 GB Mac. Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

Fits instead: gpt-oss 20B (11.6 GB)

The VRAM budget

weights 60.5 GB
Weights 60.5 GB KV cache @ 8K 0.29 GB Runtime overhead 0.6 GB Over budget 43.0 GB past 18.4 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
MXFP4 60.5 GB 61.4 GB — ~7.0 Reference 43.0 GB over

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (128 tokens, 1 global in 2), which is why its cache barely grows with context.

How to run it

terminal
$ llama-server \
    -hf openai/gpt-oss-120b:MXFP4 \
    -c 8192 -ngl 10

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 60.5 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on an RTX 4000 Ada

ModelTotalTok/s est.
gpt-oss 20B
20.9B MoE · same family
MXFP4 · 11.6 GB 45 Runs great
Mistral Small 4 119B-A6B
119B MoE · similar size
Q4_K_M · 67.7 GB ~4.9 CPU offload
Nemotron 3 Super 120B-A12B
124B MoE · similar size
Q4_K_M · 70.4 GB ~2.6 CPU offload
Ling 3.0 Flash 124B-A5B
124B MoE · similar size
Q4_K_M · 70.4 GB ~6.2 CPU offload
Qwen3.5 122B-A10B
125B MoE · similar size
Q4_K_M · 71.1 GB ~3.1 CPU offload
Llama 4 Scout 109B-A17B
109B MoE · similar size
Q4_K_M · 63.4 GB ~1.9 CPU offload
Qwen3.8-Flash-Next 180B-A6B
180B MoE · similar size
Q4_K_M · 102.0 GB ~4.8 CPU offload

Every model against the RTX 4000 Ada →

gpt-oss 120B on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
61.4 GB ~5.7 CPU offload
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
61.4 GB ~6.1 CPU offload
M4 · 16 GB
10.7 GB usable · 120 GB/s
61.4 GB — Won't fit
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
61.4 GB ~7.9 CPU offload
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
61.4 GB ~7.9 CPU offload
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
61.4 GB — Won't fit
H100 SXM
78.4 GB usable · 3350 GB/s
61.4 GB 295 Runs great
DGX Spark · 128 GB
96.0 GB usable · 273 GB/s
61.4 GB 27 Runs great

All 114 devices, by tier →