community pool / model answer

Qwen3.5-9B-FP8-dynamic

chat · RedHatAI/Qwen3.5-9B-FP8-dynamic

123 tok/s · model median · 2 runs in the model record.

Answer

147.8tok/s

reportedArc Pro B70 32GB ×28-bit · 1 run · decodebest cell in the pool

  1. Arc Pro B70 32GB ×2147.8 tok/sreported1 run · peak VRAM no data yet · 64 GB rig memory
  2. Arc Pro B70 32GB98.1 tok/sreported1 run · peak VRAM no data yet · 32 GB rig memory
Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B70 32GB8-bit98.1 tok/sreported75- GB1,024vllm
Arc Pro B70 32GB ×28-bit147.8 tok/sreported50- GB1,024vllm

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.