▲ local intelligence

can it run

Can the NVIDIA GTX 1080 run qwen2.5vl 7B?

FULL

5.8 GB needed of 8 GB usable — headroom for context

8GB · Pascal · model file 5.6GB · 7B

~16–29 tok/s

computed roofline: 320 GB/s × 0.28–0.50 efficiency window / 5.6GB (Q4_K_M) — bands, never points.

Run qwen2.5vl 7B on instead

→ Tesla M60 (16GB, used) — cheapest catalog machine that runs qwen2.5vl 7B fully ($35)

→ rtx-5060-ti-16 — the next step up from NVIDIA GTX 1080

→ every model the NVIDIA GTX 1080 can run — full list

→explore the full catalog91 machines indexed · live prices · what each one can run
2026-10-01 · ← all models on the NVIDIA GTX 1080 · fit = weights + 4k KV vs usable memory