ghost_gpu
@ghost-gpuRegular
Two used GPUs, one question: what's still worth running at home.
- Used hardware
- Quantization
Poland · English, Polish · Europe/Warsaw
Runs
11 verified
ModelWhereSpeed
MiniCPM5-2B Q4_K_MSep 12, 2026RTX 3090 · 24 GB248.9 tok/s
Llama-3.1-8B-Instruct Q4_K_MSep 5, 2026RX 7900 XTX · 24 GBNot counted
Qwen3.5-2B Q4_K_MAug 31, 2026RTX 3090 · 24 GB259.9 tok/s- Qwen3-4B Q4_K_MAug 25, 2026RX 7900 XTX · 24 GB215.1 tok/s
gemma-3-4b-it Q4_K_MAug 19, 2026RTX 3090 · 24 GB253.9 tok/s- gemma-3-12b-it Q4_K_MAug 14, 2026RX 7900 XTX · 24 GB67.0 tok/s
- Qwen3-Coder-30B-A3B-Instruct Q4_K_MAug 6, 2026RTX 3090 · 24 GB206.5 tok/s
- Qwen3.5-2B Q4_K_MAug 1, 2026RX 7900 XTX · 24 GB216.9 tok/s
PassedFailedNot supported
Accounts, metrics and works carry separate evidence. Updated Sep 12, 2026. Public JSON