Timeline
NVIDIA announced NVFP4 4-bit precision format for Blackwell GPUs integrated into MaxText.
NVFP4 achieves 1.8x training speedup over FP8 on GPT-3 175B model with no accuracy loss claimed for models up to 70B.
Qualified HBM4 from Samsung, SK Hynix, and Micron for Vera Rubin
Nvidia launches Cosmos 3 at CVPR
Launched physical AI agent skills, Alpamayo 2 Super, Cosmos 3, and other tools at CVPR
Nvidia acquired Kumo AI for over $400 million
HuggingFace launches Daily Papers SKILL.md for AI agents to read, search, and fetch research papers.
Launched 'Kernels' hub on its platform for sharing and discovering optimized GPU code.
Announced completion of a large-scale project converting 27,000 arXiv papers from PDF to Markdown using an open 5B model.
Julien Chaumond, CTO of Hugging Face, announced the Safetensors transfer via social media.
Ecosystem
Hugging Face
Nvidia
Evidence (6 articles)
NVIDIA Lyra 2.0 Launches on Hugging Face for Persistent 3D World Generation
Apr 18, 2026NVIDIA's Kimi-K2.5 Eagle Head: Supercharging Moonshot's Reasoning with Speculative Decoding
Mar 12, 2026NVIDIA Releases Brain MRI Generation Model on Hugging Face: 3D Latent Diffusion for T1, FLAIR, T2, and SWI Scans
Mar 15, 2026NVIDIA's DiffiT: A New Vision Transformer Architecture Sets Diffusion Model Benchmark
Mar 9, 2026Google Open-Sources DiffusionGemma, 26B Model Hits 1K Tokens/Sec on H100
Jun 10, 2026NVIDIA Releases NVPanoptix-3D on Hugging Face: Single-Image 3D Indoor Scene Reconstruction
Mar 24, 2026