AI BriefingHuggingFace Blog news
The latest AI and developer news from HuggingFace Blog, with short summaries.
Visit website →Google unveils multimodal-capable Gemma 3
Google has released Gemma 3, a new open model supporting multimodality and a 128k context.
OlympicCoder and New Coding Datasets Released
The Open R1 team has released OlympicCoder, a coding model that outperforms Claude 3.7 Sonnet, along with new datasets.
World's Largest Autonomous Driving Dataset L2D Released
Hugging Face's LeRobot team and Yaak have released L2D, the world's largest open-source autonomous driving dataset at 90TB in scale.
Cohere Unveils Multilingual VLM 'Aya Vision'
Cohere For AI has released the 'Aya Vision' series, an open-weight multimodal model supporting 23 languages.
Hugging Face Partners with JFrog to Strengthen AI Security
Hugging Face is adopting JFrog's scanner to enhance detection of malicious code within model weights.
Implementing Agent Observability with Arize Phoenix
This covers how to trace and evaluate the behavior of AI agents built with smolagents in real time using Arize Phoenix.
HuggingFace Supports Vaani, India's Multilingual Dataset
HuggingFace partnered with IISc to release Vaani, a large-scale multimodal dataset covering India's diverse languages and dialects.
HuggingFace Unveils FastRTC for Real-Time AI
HuggingFace has released the FastRTC library, making it easy to develop real-time audio/video AI apps in Python.
Hugging Face Supports Remote VAE Decoding
It supports offloading decoding remotely to reduce the memory burden of the VAE decoder during high-resolution image/video generation.
Google Unveils SigLIP 2 Vision-Language Encoder
Google has released SigLIP 2, a new vision-language encoder with improved performance and multilingual capabilities.
HuggingFace Releases SmolVLM2, an Ultra-Lightweight Video Understanding Model
HuggingFace has released the SmolVLM2 series of ultra-lightweight multimodal models with video understanding capabilities.
Google Releases PaliGemma 2 Mix
Google has launched the PaliGemma 2 Mix models, optimized for a variety of vision-language tasks.
Hugging Face Adds 3 New Inference Partners
Hugging Face has added Hyperbolic, Nebius AI Studio, and Novita as new serverless inference providers.
HF Improves Math Evaluation on LLM Leaderboard
Hugging Face resolved math evaluation errors on the Open LLM Leaderboard by introducing Math-Verify and recalculated the rankings.
Fireworks.ai added as an HF Inference Provider
Fireworks.ai has been added as a serverless inference provider on the Hugging Face Hub, supporting major LLMs.
Cost Analysis for Classification and Embedding at 1 Billion Requests
It presents a methodology for optimizing hardware-specific costs and latency for large-scale classification and embedding workloads.
HuggingFace Releases Video Dataset Building Toolkit
It introduces a 3-stage dataset building pipeline and open-source toolkit for training video generation models.
Hugging Face Improves Data Transfer Speed by 2-3x
Hugging Face introduced a block-based data management approach to improve model and dataset transfer speeds.
OpenR1-Math-220k Dataset Released
HuggingFace has released OpenR1-Math-220k, a mathematical reasoning dataset aimed at reproducing DeepSeek R1's reasoning capabilities.
Open Arabic LLM Leaderboard 2 Released
The Open Arabic LLM Leaderboard 2 has been released to transparently evaluate the performance of Arabic LLMs.