Deep Learning
2 articles

Research & Breakthroughs
Researchers benchmark instruction-tuned LLMs using FP8, GPTQ, and SmoothQuant
FP8 quantization slashes the memory footprint of 70B-class open-weight models, maintaining accuracy within 0.
David Katzman·May 18, 2026

Research & Breakthroughs
NVIDIA unveils NVFP4 4-bit pretraining methodology for efficient AI
NVIDIA has successfully pretrained a 12-billion-parameter Mamba-Transformer model on 10 trillion tokens using a new 4-bit precision method.
David Katzman·May 18, 2026