A 4-bit model just beat its full-precision original: meet QAH
Compression usually costs accuracy. Quantization-Aware Healing flips that: a GPT-OSS 120B model squeezed to 60B parameters and MXFP4 weights beats its own bfloat16 original on 7 of 9 benchmarks. The trick is distilling from the pre-compression model as teacher, not the degraded recovered checkpoint. Smaller, cheaper to run, and more accurate.