can it run
21.0 of 20 GB usable — barely over; real with q8 KV cache (halves KV) or shorter context
20GB · RDNA3 · model file 20GB · distill 32B
computed roofline: 800 GB/s × 0.50–0.75 efficiency window / 20.0GB (Q4_K_M) — bands, never points.
→ Tesla V100 32GB (used) — cheapest catalog machine that runs DEEPSEEK-R1 distill 32B fully ($700)
→ rx-7900-xtx — the next step up from AMD Radeon RX 7900 XT
→ every model the AMD Radeon RX 7900 XT can run — full list
→explore the full catalog91 machines indexed · live prices · what each one can run