AI BriefingHuggingFace Blog news
The latest AI and developer news from HuggingFace Blog, with short summaries.
Visit website →Together AI Supports Fine-Tuning of Hugging Face Models
Various LLMs on the Hugging Face Hub can now be easily fine-tuned through Together AI's infrastructure.
HuggingFace Develops Jupyter Agent
HuggingFace has developed Jupyter Agent, which performs data science tasks by executing code within a notebook environment.
mmBERT Released, Supporting 1,800 Languages
mmBERT, a next-generation multilingual encoder based on the ModernBERT architecture and supporting over 1,800 languages, has been released.
Google Unveils Lightweight Multilingual Embedding Model
Google has released EmbeddingGemma, a 308M-scale multilingual embedding model optimized for on-device environments.
SAIR Releases 5.24 Million-Entry Dataset
SandboxAQ has released SAIR, a dataset of 5.24 million 3D protein-ligand structures for predicting drug efficacy.
ZeroGPU Spaces Achieves 1.8x Performance Boost with AoT Compilation
This introduces a method to speed up model inference on ZeroGPU Spaces by up to 1.8x using PyTorch's AoT compilation.
NVIDIA, 6 Million-Example Multilingual Reasoning Dataset
NVIDIA has released the Nemotron multilingual reasoning dataset, comprising 6 million examples across 5 languages, along with the Nemotron Nano 2 9B model.
Claude Generates Images via Hugging Face Integration
By connecting the Hugging Face MCP server to Claude, users can directly use the latest image generation models.
How to Use MCP-Based AI Research Automation Tools
This introduces a way to connect AI agents directly to research tools such as arXiv and GitHub using the Model Context Protocol (MCP).
Hugging Face Releases CUDA Kernel Build Tool
Hugging Face has released the kernel-builder library, which supports everything from developing to deploying custom CUDA kernels.
Kimina-Prover-RL: Open-Source Training Pipeline for Lean 4 Theorem Proving Released
The Kimina-Prover-RL pipeline, which performs Lean 4 theorem proving using DeepSeek-R1-style reasoning, has been released along with two small models.
Arm KleidiAI Now Built into ExecuTorch 0.7
Arm KleidiAI is now enabled by default in ExecuTorch 0.7, greatly improving LLM inference performance on mobile and edge devices.
Arm Launches NSS, AI Upscaling for Mobile
Arm has unveiled Neural Super Sampling (NSS), an AI-based upscaling solution that optimizes mobile GPU performance.
New LLM Evaluation Benchmark for Filipino Languages Released
FilBench, a new benchmark for evaluating comprehension and generation abilities in Filipino languages (Tagalog, Cebuano), has been released.
AI Agent Benchmark TextQuests Released
TextQuests, a text-based game benchmark for evaluating LLMs' reasoning and exploration abilities as agents, has been announced.
Hugging Face Supports ND-Parallel
Hugging Face's Accelerate library has launched ND-Parallel, a feature that supports efficient multi-GPU training by combining various parallelization strategies.
Hugging Face Unveils AI Sheets
Hugging Face has launched AI Sheets, a no-code tool that lets you build and transform datasets using open models.
TRL Expands Support for VLM Alignment Techniques
HuggingFace's TRL library now supports the latest alignment techniques such as MPO and GRPO for VLMs.
OpenAI Releases Open Source Model 'GPT OSS'
OpenAI has released the GPT OSS series, an open-weight model lineup under the Apache 2.0 license.
NVIDIA AI-Q Ranks #1 on DeepResearch Bench
NVIDIA's AI-Q Blueprint, leveraging Llama Nemotron models, ranked #1 among open source stacks on DeepResearch Bench.