community pool / model answer

Qwen3.5-9B-quantized.w4a16

chat · RedHatAI/Qwen3.5-9B-quantized.w4a16

124.1 tok/s · model median · 3 runs in the model record.

Answer

172.3tok/s

reportedArc Pro B70 32GB ×24-bit · 1 run · decodebest cell in the pool

  1. Arc Pro B70 32GB ×2172.3 tok/sreported1 run · peak VRAM no data yet · 64 GB rig memory
  2. Arc Pro B70 32GB118.7 tok/sreported2 runs · peak VRAM no data yet · 32 GB rig memory
Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
Arc Pro B70 32GB4-bit118.7 tok/sreported66- GB1,024vllm
Arc Pro B70 32GB ×24-bit172.3 tok/sreported45- 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.