List your company

Zero-shot Evaluation on Non-reasoning tasks

70.8Accuracy (Zero-shot Non-reasoning)FP16
66.43267.56668.769.834Nov 13, 2025

Evaluation results

MethodLinks
FP16Quantization Format=FP162025.1170.8—
FlatQuantQuantization Format=INT4, Bit-width=W4A4, Weight Quantization Method=GPTQ2025.1169.4—
ParoQuantQuantization Format=INT4, Bit-width=W4A4, Weight Quantization Method=GPTQ2025.1167.7—
SpinQuantQuantization Format=INT4, Bit-width=W4A4, Weight Quantization Method=GPTQ2025.1166.6—
FP16Format=FP16, Quantization=None2025.11—70.8
MR-GPTQFormat=MXFP4, Quantization=W4A4, Weight Quantization=GPTQ2025.11—68.4
ParoQuantFormat=MXFP4, Quantization=W4A4, Weight Quantization=GPTQ2025.11—69.6