Qwen3-4B-Thinking-2507-llmc-gptq-calib-c4-en-train-w4a16-g128-n256-s1024

• method: gptq
• weight format: W4A16
• group size: 128
• calibration dataset: allenai/c4
• calibration dataset config: en
• calibration split: train
• calibration samples: 256
• max sequence length: 1024