community pool / model answer
Qwen3.6-35B-A3B-FP8
chat · Qwen/Qwen3.6-35B-A3B-FP8
161.7 tok/s · model median · 86 runs in the model record.
| Reference rig | Quantization | Median result | Basis | TTFT (ms) | Peak VRAM | Max context | Engines |
|---|---|---|---|---|---|---|---|
| B200 192GB | 8-bit | 220.4 tok/s | reported | 28 | 161.3 GB | 32,768 | vllm |
| CMP 170HX 64GB | 8-bit | 155.7 tok/s | measured | 1,868 | 55.8 GB | 131,335 | vllm |
| GB10 Grace Blackwell 121GB | 8-bit | 54.1 tok/s | reported | - | - GB | 128,000 | vllm |
| H100 80GB | 8-bit | 337.1 tok/s | reported | 58 | 71.6 GB | 32,768 | vllm |
| H200 SXM 141GB | 8-bit | 261.5 tok/s | reported | 48 | 126.7 GB | 32,768 | vllm |
| RTX 3090 24GB ×2 | 8-bit | 140 tok/s | reported | 149 | 22.2 GB | 32,768 | vllm |
| RTX PRO 6000 Blackwell 96GB | 8-bit | 253.7 tok/s | reported | 229 | 38.4 GB | 4,096 | 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.