← All categories
Single GPU
One card in an ordinary desktop. The card's VRAM is the whole budget for weights and context.
| Machine | Memory for models | Bandwidth | Price | Largest at 4-bit | gpt-oss-20b |
|---|---|---|---|---|---|
| 1x RTX 3090 (used)Single GPU | 24 GB | 936 GB/s | ~$2,944 + unpriced partsused GPU price not included | Qwen3.6 35B-A3B | 162 tok/s |
| 1x RTX 5090Single GPU | 32 GB | 1,792 GB/s | ~$3,234 + unpriced partsGPU price not included | Nemotron 3.5 Lightning 30B-A3B | 290 tok/s |
| 1x RTX PRO 6000 BlackwellSingle GPU | 96 GB | 1,792 GB/s | ~$20,833sum of parts | Nemotron 3 Super 120B-A12B | 287 tok/s |