OpenAI · 117B total / 5.1B active · MoE
GPT-OSS 120B needs about 67 GB at MXFP4 (native). 11 of the 17 priced rigs we track can hold it. Here's the cheapest, the best value and the fastest. Prices & stock checked 2026-10-23. Prices change daily and some retailers block our checker, so always confirm on the store page.
Ryzen AI Max+ 395, 128GB
Cheapest that fits. Lowest price we've seen on any tracked rig that holds GPT-OSS 120B at 4-bit (67 of 96 GB usable).
Ryzen AI Max+ 395, 128GB
Best value. Lowest price per token/sec among complete systems where it fits comfortably: about $63 per tok/s at 4-bit. Same machine as the cheapest pick.
96GB card
Fastest. Most memory bandwidth of the rigs that fit it (1,792 GB/s), so the quickest decode: ~231 tok/s at 4-bit.
GPU price is card only: you still need a host PC.
Check priceUse MXFP4 (native) on most rigs: it needs ~67 GB. Q8_0 would need ~132 GB, which doesn't fit comfortably at a usable speed on anything we track. GPT-OSS 120B ships natively in MXFP4 (native), so that's the format to grab.
| Machine | Price | 4-bit fit & speed | Suggested quant | $ per tok/s |
|---|---|---|---|---|
| Framework Desktop DIYStrix Halo | $3,449 | Fits~55 tok/s | MXFP4 (native) | $63 |
| GMKtec EVO-X2Strix Halo | $3,500 | Fits~55 tok/s | MXFP4 (native) | $64 |
| Minisforum MS-S1 MaxStrix Halo | $3,919 | Fits~55 tok/s | MXFP4 (native) | $72 |
| Beelink GTR9 ProStrix Halo | $4,349 | Fits~55 tok/s | MXFP4 (native) | $80 |
| Mac Studio M5 UltraMac Studio | $5,499 | Tight fit~109 tok/s | MXFP4 (native) | $51 |
| Lenovo ThinkStation PGXDGX Spark / GB10 | $5,859 | Fits~61 tok/s | MXFP4 (native) | $95 |
| ASUS Ascent GX10DGX Spark / GB10 | $5,999 | Fits~61 tok/s | MXFP4 (native) | $98 |
| NVIDIA DGX SparkDGX Spark / GB10 | $6,950 | Fits~61 tok/s | MXFP4 (native) | $113 |
| HP ZGX Nano G1nDGX Spark / GB10 | $7,699 | Fits~61 tok/s | MXFP4 (native) | $125 |
| Gigabyte AI TOP ATOMDGX Spark / GB10 | $8,499 | Fits~61 tok/s | MXFP4 (native) | $138 |
| PNY NVIDIA RTX PRO 6000 BlackwellRTX PRO 6000 (card only, needs a host PC) | $17,899 | Fits~231 tok/s | MXFP4 (native) | $77 |
Sorted by lowest price. Speeds are estimated single-user decode at ~8K context. Full fit table, including rigs that don't fit: GPT-OSS 120B memory requirements. How we estimate
About 67 GB at MXFP4 (native) and 132 GB at Q8_0, including runtime overhead and ~8K tokens of context for one user. Longer context needs more.
Of the rigs we track, the Framework Desktop DIY at $3,449, estimated ~55 tok/s at 4-bit. Prices checked 2026-10-23.
MXFP4 (native) is the default: about 67 GB. Q8_0 needs about 132 GB; use it only if you have the room and still get usable speed.
It has 5.1B active parameters per token, so speed is set by memory bandwidth: from ~55 tok/s on the slowest rig that fits to ~231 tok/s on the PNY NVIDIA RTX PRO 6000 Blackwell. Estimates, not benchmarks.
One short email a week: what got cheaper, what's back in stock, and which new open models your machine can run. Unsubscribe anytime.
