community pool / model answer
Qwen3-14B
chat · Qwen/Qwen3-14B
33.7 tok/s · model median · 20 runs in the model record.
| Reference rig | Quantization | Median result | Basis | TTFT (ms) | Peak VRAM | Max context | Engines |
|---|---|---|---|---|---|---|---|
| M1 Max 64GB | 4-bit | 24.8 tok/s | reported | 51 | - GB | 2,048 | ollama |
| RTX 3060 12GB | 4-bit | 29 tok/s | measured | 2,553 | 8.7 GB | 65,536 | llama.cpp, ollama |
| RTX 3080 12GB | 4-bit | 75 tok/s | reported | - | 8.7 GB | 2,048 | llama.cpp |
| RTX 3090 24GB | 4-bit | 42.9 tok/s | measured | 2,655 | 0 GB | 131,072 | llama.cpp |
| RTX3060+RTX2080-pooled-20GB 20GB | 4-bit | 35 tok/s | reported | - | 8.7 GB | 2,048 | ollama |
Context tested
The table above is the tested context for each exact rig and quantization cell. Empty fields remain empty rather than being filled with an estimate.