AI BriefingHuggingFace Blog news
The latest AI and developer news from HuggingFace Blog, with short summaries.
Visit website →DABstep Benchmark for Data Agents Released
Adyen and Hugging Face have released the DABstep benchmark, which evaluates the capabilities of data analysis agents.
π0 Robot Model Released on LeRobot
Physical Intelligence's general-purpose robot control model π0 has been integrated into Hugging Face LeRobot.
HuggingFace Releases Open-Source DeepResearch
HuggingFace has released an open-source framework to reproduce OpenAI's Deep Research feature.
HuggingFace Shares Open-R1 Project Progress
HuggingFace announced the first update on the Open-R1 project, which aims to reproduce DeepSeek-R1's training pipeline and dataset.
Guide to Reproducing DeepSeek R1's 'Aha Moment'
It explains how to reproduce DeepSeek R1's reasoning capability in a small model through the GRPO algorithm.
HuggingFace Summarizes 2024 AI Art Technology
It summarized the major technical turning points and open-source model achievements in AI image and video generation in 2024.
AWS-based DeepSeek-R1 Deployment Guide
Hugging Face and AWS have partnered to guide the deployment and fine-tuning of the DeepSeek-R1 model.
HuggingFace unveils Open-R1 project
HuggingFace has launched the Open-R1 project to reproduce DeepSeek-R1's training process.
Hugging Face Integrates Serverless Inference
Hugging Face has directly integrated major serverless inference providers into the Hub and SDK.
Diffusers: The State of Open Video Models
The Diffusers team has summarized the technical status of open video generation models and their future support plans.
smolagents adds vision capabilities with VLM support
HuggingFace's smolagents has been updated with VLM support, enabling it to process visual information.
NVIDIA Releases KV Cache Compression Toolkit
NVIDIA has released KVPress, a KV cache compression toolkit for handling long context in LLMs.
Ultra-lightweight VLM SmolVLM 256M/500M released
HuggingFace has released SmolVLM 256M and 500M, ultra-lightweight multimodal models.
HF-FriendliAI Model Deployment Integration
One-click model deployment via FriendliAI is now available on the Hugging Face Hub.
TGI Adds Multi-Backend Inference Support
Hugging Face's TGI is introducing a multi-backend architecture that supports various inference engines such as vLLM and TensorRT-LLM.
Transformers Integration Support for timm Models
Hugging Face released TimmWrapper, which enables various computer vision models from timm to be used within the transformers ecosystem.
Static Embedding Models Released, Up to 400x Faster on CPU
A static embedding model and training methodology have been released that deliver up to 400 times faster speed than existing models in CPU environments.
The Rise of AI Agents and Ethical Challenges
This analyzes the technical definition of AI agents and the safety and ethical risks that come with their increasing autonomy.
Multilingual Visual Document Retrieval Model Released
An embedding model for multilingual visual document retrieval, developed in collaboration with LlamaIndex, along with a 500K dataset, has been released.
Open LLM Leaderboard Carbon Emissions Analysis
HuggingFace analyzed the correlation between model performance and CO₂ emissions using Open LLM Leaderboard data.