Apple M3 Ultra (80-core GPU, 512GB) × gpt-oss-120b (MoE)
Can the Apple M3 Ultra (80-core GPU, 512GB) run gpt-oss-120b (MoE)?
Yes
Yes. The Apple M3 Ultra (80-core GPU, 512GB) can run gpt-oss-120b (MoE) at up to MXFP4 (native) (needs ~68.4 GB of 384 GB usable), at an estimated ~50.0 tokens/sec at MXFP4 (native).
Speed is estimated from memory bandwidth calibrated with a measured 92.14 tok/s Llama 2 7B Q4_0 result (source). No direct public benchmark of this exact pair is recorded yet.
By quantization (8K context)
| Quant | Memory needed | Fits? | Est. tok/s |
|---|---|---|---|
| MXFP4 (native) | 68.4 GB | ✅ | ~50.0 |
Fastest hardware for gpt-oss-120b (MoE)
- Apple M5 Max (40-core GPU, 128GB) — ~65.1 tok/s at MXFP4 (native)
- Apple M2 Ultra (76-core GPU, 192GB) — ~51.2 tok/s at MXFP4 (native)
- Apple M4 Max (40-core GPU, 128GB) — ~45.1 tok/s at MXFP4 (native)
- NVIDIA DGX Spark — ~31.0 tok/s at MXFP4 (native)
No suitable GPU? Rent one by the hour on Vast.ai or RunPod (affiliate links).
Get new models & benchmarks in your inbox
No spam. Unsubscribe anytime by replying to any email.