community pool / model answer
Nemotron-Cascade-2-30B-A3B
chat · nvidia/Nemotron-Cascade-2-30B-A3B
155.3 tok/s · model median · 13 runs in the model record.
| Reference rig | Quantization | Median result | Basis | TTFT (ms) | Peak VRAM | Max context | Engines |
|---|---|---|---|---|---|---|---|
| M3 Ultra 512GB | 4-bit | 118.1 tok/s | reported | 719 | - GB | 8,192 | llama.cpp, mlx |
| Radeon AI Pro R9700 32GB ×3 | 8-bit | 89.8 tok/s | reported | 9,971 | - GB | 31,620 | llama.cpp |
| RTX 3060 20GB | 4-bit | 104 tok/s | reported | - | 17.6 GB | 2,048 | llama.cpp |
| RTX 3090 24GB | 4-bit | 188.1 tok/s | measured | 1,769 | 0 GB | 131,072 | llama.cpp |
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.