Models › gpt-oss 120B (MoE)

Can I run gpt-oss 120B (MoE) in Australia?

Yes. The cheapest hardware we track that holds gpt-oss 120B (MoE) at MXFP4 with an 8k context is the Minisforum MS-S1 Max (Ryzen AI Max+ 395) 128GB at A$5,719 (Minisforum AU, checked 20 Oct 2026, 7:44 am AEST). Expect roughly 70 tokens/sec (estimate).

OpenAI · released 4 Aug 2025 older · superseded but still widely used · 3.2M HF downloads (last 30 days)

117Btotal parameters · 5.1B active (MoE)
36layers (18 full-attention, 18 sliding-window)
0.31 GBKV cache @ 8k · 8 KV heads × 64 dim
Model card ↗config.json · source of counts

GGUF builds: unsloth · bartowski · ggml-org · lmstudio-community

Memory needed by quantisation (8k context)

QuantWeightsTotal neededCheapest hardware that fits
MXFP4
62.2 GB
63.5 GB
Q8_0
124.3 GB
125.6 GB

All hardware that fits at MXFP4

HardwareNeeds / usableEst. tok/sPriceRetailer link
63.5 / 96.0 GB fits
70 estimate
A$5,719
Minisforum AU
checked 20 Oct 2026, 7:44 am AEST
63.5 / 96.0 GB fits
70 estimate
A$8,999
Mwave · not in stock online
checked 20 Oct 2026, 7:44 am AEST
63.5 / 96.0 GB fits
80 estimate
A$8,999
Mwave
checked 20 Oct 2026, 7:44 am AEST
63.5 / 72.0 GB fits
328 estimate
A$9,499
Apple AU
checked 20 Oct 2026, 7:44 am AEST
63.5 / 96.0 GB fits
82 estimate
A$9,649
Microsoft Store AU · stale
checked 19 Oct 2026, 8:18 pm AEST
63.5 / 192.0 GB fits
251 estimate
A$11,449
Apple AU
checked 20 Oct 2026, 7:44 am AEST
63.5 / 95.5 GB fits
384 estimate
A$27,599
Scorptec
checked 20 Oct 2026, 7:42 am AEST
63.5 / 96.0 GB fits
70 estimate
price unavailable

The tok/s figures are bandwidth-based estimates for this model. Where a device has a measured benchmark, the estimate is calibrated to that run. They are not measurements. Methodology.

Graphics card + system RAM (MoE experts in RAM)

Mixture-of-experts models only read a few experts per token, so llama.cpp can keep the expert weights in system RAM and everything else on the GPU (e.g. --n-cpu-moe / -ot exps=CPU). That lets a 12–32GB card run a much bigger model. All figures below are estimates (MXFP4, 8k context, dual-channel DDR5-5600). How we estimate.

Graphics cardCard priceGPU only+ 32GB RAM+ 64GB RAM+ 128GB RAM
A$419
no
no
tight
53 GB in RAM · ~29 tok/s
fits
53 GB in RAM · ~29 tok/s
A$539
no
no
tight
53 GB in RAM · ~29 tok/s
fits
53 GB in RAM · ~29 tok/s
A$699
no
no
fits
50 GB in RAM · ~30 tok/s
fits
50 GB in RAM · ~30 tok/s
A$1,099
no
no
fits
50 GB in RAM · ~31 tok/s
fits
50 GB in RAM · ~31 tok/s
A$1,099
no
no
fits
50 GB in RAM · ~32 tok/s
fits
50 GB in RAM · ~32 tok/s
A$1,169
no
no
fits
50 GB in RAM · ~32 tok/s
fits
50 GB in RAM · ~32 tok/s
A$1,249
no
no
tight
53 GB in RAM · ~30 tok/s
fits
53 GB in RAM · ~30 tok/s
A$1,699
no
no
fits
50 GB in RAM · ~33 tok/s
fits
50 GB in RAM · ~33 tok/s
A$2,299
no
no
fits
50 GB in RAM · ~33 tok/s
fits
50 GB in RAM · ~33 tok/s
A$2,299
no
no
fits
35 GB in RAM · ~43 tok/s
fits
35 GB in RAM · ~43 tok/s
A$3,599
no
no
fits
42 GB in RAM · ~37 tok/s
fits
42 GB in RAM · ~37 tok/s
A$10,999
no
no
fits
35 GB in RAM · ~46 tok/s
fits
35 GB in RAM · ~46 tok/s
A$27,599
fits
~384 tok/s
fits in VRAM
fits in VRAM
fits in VRAM
n/a
no
no
fits
50 GB in RAM · ~29 tok/s
fits
50 GB in RAM · ~29 tok/s
n/a
no
no
fits
42 GB in RAM · ~37 tok/s
fits
42 GB in RAM · ~37 tok/s
n/a
no
no
fits
42 GB in RAM · ~38 tok/s
fits
42 GB in RAM · ~38 tok/s
n/a
no
no
fits
42 GB in RAM · ~38 tok/s
fits
42 GB in RAM · ~38 tok/s

DDR5 kit prices today: 32GB: A$610 (Mwave, checked 20 Oct 2026, 7:44 am AEST) · 64GB: A$1,499 (Mwave, checked 20 Oct 2026, 7:44 am AEST) · 128GB: A$2,511 (Mwave, not in stock online, checked 20 Oct 2026, 7:44 am AEST). Card + kit only; the rest of the PC isn't included. RAM-offload speeds are estimates and can be optimistic, so check community results below.

Community measured

No reader results yet. Be the first to submit one →

Try other quantisations and context lengths in the Can I run it? tool → · Submit your own gpt-oss 120B (MoE) benchmark →

Get the weekly "best value AI rig in AUD" email

Price drops on cards and boxes that run local models, plus new models and what you need to run them. One email a week.