community pool / model answer

Qwen3.8-27B-EXL3-3.5bpw

chat · Mia-AiLab/Qwen3.8-27B-EXL3-3.5bpw

115.1 tok/s · model median · 5 runs in the model record.

Answer

115.1tok/s

measuredRTX 3090 24GB3-bit · 5 runs · decodebest cell in the pool

  1. RTX 3090 24GB115.1 tok/smeasured5 runs · peak 17.3 GB · fits 24 GB rig memory
Reference rigQuantizationMedian resultBasisTTFT (ms)Peak VRAMMax contextEngines
RTX 3090 24GB3-bit115.1 tok/smeasured59317.3 GB24,576exllamav3

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.