Hugging Face· PlatformResearch paper

Quantization-Aware Healing: a compressed, 4-bit model that outperforms its full-precision original

Introduces Quantization-Aware Healing (QAH), a method that distills from the original pre-compression model to recover capabilities lost during structural compression and quantization. Applied to a GPT-OSS 120B model compressed to 60B and quantized to MXFP4, it outperforms its own full-precision version on 7 of 9 benchmarks.

From the source

applied to a GPT-OSS 120B model compressed to 60B parameters and quantized to MXFP4, it produces a model that beats its own full-precision (bfloat16) version on 7 of 9 benchmarks.

huggingface.co

Announced

25 Aug 2026 11:39 UTC

Detected

25 Aug 2026 11:45 UTC (+5m after)