can it run
Only with compromises. 8GB is below the 9GB working set — expect a lower quant, offloading, or heavy swap. Possible, not pleasant.
weights 7.1GB + KV cache 0.62GB at 4k context vs ~8GB usable → tight. 7.7 of 8 GB usable — keep context modest
computed roofline: 192.3 GB/s × 0.28–0.50 efficiency window / 7.1GB (Q4_K_M) — bands, never points.
| model file | 7.1GB (GGUF Q4_K_M) |
| minimum memory | 9GB working set |
| this machine | 8GB GDDR5, 192 GB/s |
| params | 12B |
| license | apache-2.0 |
| price | $100 |
easy run: ollama pull mistral-nemo:12b
runs with: ollama · llama.cpp · koboldcpp · lm studio
→ Tesla M60 (16GB, used) — cheapest machine that runs it ($35)
→ every model the Tesla P4 (8GB, used) can run — full list
→ tesla-p100-16gb — the next step up from this machine's GPU