Z.ai and NVIDIA engineers trained GLM-5.2, a 744B-parameter mixture-of-experts model quantized to 4-bit NVFP4, with reinforcement learning to play Super Mario Bros., overcoming arithmetic, distributed…
A developer's analysis of low-precision AI formats in mid-2026 shows that FP8 and FP4 have become essential for efficient large-scale AI training and inference. The report details how FP8 formats (E4M…
NVIDIA detailed its new "Rubin" GPU architecture, which delivers 50 PetaFLOPS at sparse NVFP4 operations, features 224 streaming multiprocessors, 288 GB of HBM4 memory, 336 billion transistors, and 89…
NVIDIA's Rubin GPU, the core of the Vera Rubin platform, delivers up to 10x more agentic throughput per unit of energy than the Blackwell architecture, according to NVIDIA. The GPU features 336 billio…
NVIDIA released a guide showing how to optimize transformer-based models for low-precision training using Hopper and Blackwell GPUs, focusing on FP8 and NVFP4 formats. The method translates model conf…
NVIDIA released BioNeMo Recipes, a set of training recipes that use Low-Rank Adaptation (LoRA) to fine-tune large biological foundation models like ESM2-3B and Evo2-1B on a single workstation GPU. The…
NVIDIA introduced advanced fused MLP kernels for mixture-of-experts (MoE) models, built with the CuTe DSL, delivering 1.3x–2x kernel-level speedups and enabling sync-free MoE execution. The optimizati…