Hugging Face's LeRobot v0.4.0 Aims to Accelerate Open-Source Robotics
The new version of Hugging Face's open-source robot learning library is positioned as a major step for accessible robotics development.
AI & ML
Frontier models, research results, agents, and developer tooling.
BigCodeArena Judges AI Code by Running It
Hugging Face’s BigCodeArena evaluates code generation models through actual execution, shifting the focus from static scoring to functional results.
NVIDIA's Nemotron-Personas-India Dataset Targets Sovereign AI
The synthetic-persona dataset is designed to help developers build AI systems for Indian languages and local contexts.
Synthetic Photo Albums: Coherent and Private by Design
Google Research proposes a hierarchical method for generating synthetic photo albums that stay narratively coherent while protecting privacy.
AI Sheets Brings Image Analysis to Spreadsheets
Hugging Face's new tool integrates vision capabilities directly into spreadsheet workflows, enabling image-based data processing.
Arm to Showcase PyTorch Work at Upcoming Conference
A Hugging Face blog post announces Arm's participation in PyTorch Conference, inviting the community to join.
Depth-Pruned Draft Models Speed Up Qwen3-8B Agent on Intel Core Ultra
A Hugging Face blog shows how layer-pruned draft models could make an 8B-parameter agent more practical on local Intel hardware.
Running Vision Language Models on Intel CPUs: A 3-Step Guide
Hugging Face's walkthrough shows developers how to use OpenVINO to run vision language models on Intel CPUs.
dots.ocr Brings State-of-the-Art OCR to Apple Core ML
A Hugging Face blog post highlights dots.ocr as a state-of-the-art optical character recognition model built for Apple's Core ML framework.
Open Model OCR Pipelines Gain Traction
A Hugging Face walkthrough shows how open-weight models are now a practical alternative to closed OCR APIs for production document processing.
Hugging Face Proposes RTEB as New Benchmark for Retrieval Evaluation
The new standard aims to bring more rigorous and comparable evaluation to retrieval systems.
Hugging Face Blog Tackles AI for Food Allergies
A new post in Hugging Face's science series examines how machine learning could be applied to food allergy challenges.
AlphaFold's Five-Year Impact on Biological Discovery
A new Google DeepMind retrospective highlights how AlphaFold has accelerated science and fueled a global wave of biological discovery.
DeepMind's D4RT speeds up 4D scene reconstruction and tracking by 300x
A new unified method from Google DeepMind dramatically accelerates how AI models reconstruct and track moving 3D scenes over time.
Hugging Face Marks One Year Since DeepSeek's Breakout
A new retrospective from Hugging Face assesses how the AI field changed after DeepSeek's emergence.
OVHcloud Now Available on Hugging Face Inference Providers
The announcement adds OVHcloud as an infrastructure choice for running models through Hugging Face's hosted inference service.
DeepMind Veo 3.1 Adds Vertical Video and More Control
Google DeepMind's updated Veo model aims for more consistent, natural-looking clips and now supports vertical formats.
Gemma Scope 2 brings interpretability tools to all Gemma 3 models
Google DeepMind's open interpretability suite now covers the entire Gemma 3 family, giving safety researchers a clearer view into model internals.