AI BriefingHuggingFace Blog news
The latest AI and developer news from HuggingFace Blog, with short summaries.
Visit website →Ettin Reranker: 6 Models Released
Hugging Face released 6 Ettin-based rerankers along with the training recipe.
PaddleOCR 3.5 Adds Transformers Backend Support
PaddleOCR 3.5 has added Transformers backend support.
Open Agent Leaderboard Unveiled
Hugging Face has released a leaderboard that compares agent systems' performance and cost together.
Granite Multilingual Embedding R2 Released
IBM Granite released 97M and 311M multilingual embedding R2 models under Apache 2.0.
AWS Organizes Foundation Model Infrastructure
AWS has organized its infrastructure stack for foundation model training and inference, along with the latest GPU/EFA configurations.
OncoAgent: A Multi-Agent Framework for Oncology
OncoAgent, an open-source multi-agent framework that supports decision-making based on oncology guidelines while protecting personal information, has been released.
EMO, Document-Level MoE Pretraining
EMO has been released, maintaining performance even with only some experts used, thanks to document-level routing.
Open ASR Leaderboard Adds Private Data
Hugging Face's Open ASR Leaderboard has added private datasets to prevent benchmaxxing.
DeepInfra Joins HF Inference Providers
DeepInfra has been added to Hugging Face's Inference Providers.
NVIDIA Unveils Ultrasound Raw2Insights-US
NVIDIA has unveiled Raw2Insights-US, which estimates speed of sound from raw ultrasound data.
Building Web Apps with Privacy Filter
Three web apps applying OpenAI's Privacy Filter and how to use gradio.Server were introduced.
DeepSeek-V4, 1M-token context
DeepSeek-V4 unveiled a 1M-token context and agent optimizations.
Transformers.js for Chrome Extensions
Outlines the structure for attaching local AI to Manifest V3 Chrome extensions using Transformers.js.
Gemma 4 VLA Running on Jetson
A procedure for running Gemma 4 VLA locally on a Jetson Orin Nano Super was shared.
Cleaning Up Arabic Benchmarks
QIMMA first validates Arabic benchmarks before evaluating LLMs on them.
The Age of AI Agents: The Importance of Openness in Cybersecurity
Through the case of Anthropic's Mythos, this piece analyzes the advancement of AI agent-based cybersecurity and the importance of an open ecosystem.
Training Shopping Bots Too
Results from training a shopping agent with RL across 8 verifiable e-commerce environments.
VDR Performance Reversal
A summary of how to directly train multimodal embedding and reranker models using Sentence Transformers.
Mapping Agent Failures
VAKRA exposes agent reasoning failures in enterprise-style API and document tasks.
World Models on Home GPUs
Waypoint-1.5 extends real-time world models for consumer GPUs with 720p/60FPS and 360p tiers.