Qwen3-0.6B-3bit-awq-obq

Qwen3-0.6B quantized to 3-bit (group_size 256, symmetric), 8-bit embedding, ~4.40 bits/weight.

Method: AWQ smoothing + OBQ (GGN + KL-teacher correction). Calibration: gitarist/calibration-generic.

wikitext-2 PPL 45.98, mean KL to fp16 0.965 (fp16 ref PPL 20.96).

Weights are provided dequantized in fp16 for direct loading; quantization is baked in — evaluate as-is.