community pool / model answer
Qwen3.8-27B-NVFP4
chat · unsloth/Qwen3.8-27B-NVFP4
89.5 tok/s · model median · 15 runs in the model record.
| Reference rig | Quantization | Median result | Basis | TTFT (ms) | Peak VRAM | Max context | Engines |
|---|---|---|---|---|---|---|---|
| GB10 Grace Blackwell 128GB | 4-bit | 19.1 tok/s | measured | 197 | - GB | 131,072 | sglang, vllm |
| GB10 Grace Blackwell 256GB | 4-bit | 17.4 tok/s | reported | 168 | - GB | 2,048 | vllm |
| Radeon AI Pro R9700 32GB ×2 | 4-bit | 229.4 tok/s | measured | 68 | 57 GB | 262,144 | vllm |
| RTX 5090 32GB | 4-bit | 87.1 tok/s | measured | 135 | 30 GB | 153,664 | vllm |
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.