community pool / model answer

Qwen3.8-27B-MTP-GGUF

chat · Jackrong/Qwen3.8-27B-MTP-GGUF

36.6 tok/s · model median · 235 runs in the model record.

Answer

42tok/s

measuredCMP 170HX 64GB4-bit · 104 runs · decodebest cell in the pool

  1. CMP 170HX 64GB42 tok/smeasured104 runs · peak VRAM no data yet · 64 GB rig memory · 2 cells
Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
CMP 170HX 64GB4-bit42 tok/smeasured24,545- GB131,296llama.cpp
CMP 170HX 64GB8-bit32.7 tok/smeasured22,198- GB132,352llama.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.