Editing a copy of Strix Halo 128GB mini PC
Choose parts or a complete machine. Compatibility, the total and what it runs update as you go, and nothing is saved until you share it.
A complete machine needs no part checks.
Best fit:DeepSeek V4 FlashUD-Q2_K_XLat 32k, 9 tok/sest.
| Component | Selection | Price | Where | Actions |
|---|---|---|---|---|
| Machine | Framework Desktop Ryzen AI Max+ 395 128GB 128 GB unified · 256 GB/s | $3,449 | Buy | |
| Total | $3,449 | Prices checked 2 h ago | ||
What this build runs
The most precise quant that fits, for every model in the catalogue. Estimates, not measurements.
| Model | Quant | Fits | Memory | Max context | Decode @32k | Prefill | Good for |
|---|---|---|---|---|---|---|---|
| DeepSeek V4 Flash DeepSeek | UD-Q2_K_XL | fits | 93.9 GB | 55k99k with q8 cache | 9tok/sest. | 156 tok/s | – |
| Qwen 3.8 Flash Next Alibaba Qwen | UD-Q3_K_XL | fits | 85.6 GB | 256k | 3tok/sest. | 163 tok/s | – |
| MiniMax M2.7 MiniMax | UD-Q2_K_XL | fits | 78.9 GB | 101k186k with q8 cache | 7tok/sest. | 442 tok/s | – |
| Qwen 3.8 27B Alibaba Qwen | Q8_0 | fits | 30.2 GB | 256k | 5tok/sest. | 335 tok/s | – |
| Qwen 3.5 122B-A10B Alibaba Qwen | Q6_K | fits | 96.0 GB | 33k53k with q8 cache | 9tok/sest. | 786 tok/s | – |
| Qwen 3.6 27B Alibaba Qwen | Q8_0 | fits | 29.8 GB | 256k | 5tok/sest. | 335 tok/s | – |
| gpt-oss 120B OpenAI | MXFP4 | fits | 61.2 GB | 128k | 20tok/sest. | 1270 tok/s | chat |
| Qwen3 Coder Next Alibaba Qwen | Q8_0 | fits | 80.8 GB | 256k | 17tok/sest. | 1895 tok/s | chat |
| Nemotron 3 Super 120B-A12B NVIDIA | Q5_K_M | fits | 91.4 GB | 256k | 9tok/sest. | 795 tok/s | – |
| Gemma 4 31B | Q8_0 | fits | 34.7 GB | 256k | 4tok/sest. | 299 tok/s | – |
$3,449
1 item