Pick your hardware. See real measured speeds — no cloud estimates.
Real measured speeds on real consumer hardware — not marketing numbers. Select your GPU / RAM below.
| Model | Size | Speed | Best for |
|---|---|---|---|
| Ling-3.0-tiny | 4.8 GB | 19 tok/s (CPU) | Any PC, everyday chat |
| R1-0528-8B | 5.0 GB | 31.5 tok/s (GPU) | Math, deep reasoning |
| Qwen3.8-27B | 17.7 GB | 16GB+ VRAM | Coding, vision, agentic |
All speeds measured locally with llama.cpp, not cloud estimates. Your mileage varies with CPU/GPU.