can it run
14.2 of 13.5 GB usable — barely over; real with q8 KV cache (halves KV) or shorter context
~13.5GB usable · GCN5.1 (iGPU) · model file 14GB · 20B MoE
computed roofline: 51.2 GB/s × 0.25–0.80 efficiency window / 2.1GB (Q4_K_M) — bands, never points.
MoE: file is 14GB (that decides fit) but only ~3B of experts activate per token (~2.1GB streamed) — band below is the active-expert estimate, optimistic; community MoE rows run several× the naive bandwidth math
→ Tesla P40 (24GB, used) — cheapest catalog machine that runs gpt-oss 20B MoE fully ($290)
→ every model the Radeon Vega 8 (iGPU) can run — full list
→explore the full catalog84 machines indexed · live prices · what each one can run