The cheapest cataloged machine that runs QwQ 32B on the GPU is a AMD Radeon RX 7900 XTX · 24GB VRAM. Here are the most affordable rigs that fit it, smallest memory first.
| machine | best fit | verdict | speed |
|---|---|---|---|
| AMD Radeon RX 7900 XTX · 24GB VRAM | 32B Q4_K_M | ✅ Runs (tight) | 17–30 tok/s |
| NVIDIA GeForce RTX 3090 · 24GB VRAM | 32B Q4_K_M | ✅ Runs (tight) | 18–30 tok/s |
| NVIDIA GeForce RTX 4090 · 24GB VRAM | 32B Q4_K_M | ✅ Runs (tight) | 19–30 tok/s |
| NVIDIA GeForce RTX 5090 · 32GB VRAM | 32B Q5_K_M | ✅ Runs (tight) | 30–50 tok/s |
| Apple M3 Max · 48GB | 32B Q5_K_M | ✅ Runs comfortably | 7–12 tok/s |
| Apple M4 Max · 48GB | 32B Q5_K_M | ✅ Runs comfortably | 9–16 tok/s |
| Apple M4 Pro · 48GB | 32B Q5_K_M | ✅ Runs comfortably | 6–10 tok/s |
| AMD Radeon Pro W7900 · 48GB VRAM | 32B Q8_0 | ✅ Runs (tight) | 9–15 tok/s |
| Apple M1 Max · 64GB | 32B Q8_0 | ✅ Runs (tight) | 5–8 tok/s |
| Apple M1 Ultra · 64GB | 32B Q8_0 | ✅ Runs (tight) | 6–11 tok/s |
| Apple M2 Max · 64GB | 32B Q8_0 | ✅ Runs (tight) | 5–8 tok/s |
| Apple M2 Ultra · 64GB | 32B Q8_0 | ✅ Runs (tight) | 6–11 tok/s |
Already have a machine? See QwQ 32B across popular rigs → or check your exact rig →
Planning a whole workflow? Rigs for a coding agent · local chat · reasoning · local RAG setup.
Verdicts use the same GQA-aware fit math as the CLI — a AMD Radeon RX 7900 XTX verdict here matches what runlocal tells you on the machine.