AI BriefingHuggingFace Blog news
The latest AI and developer news from HuggingFace Blog, with short summaries.
Visit website →STEM and Code Benchmark '3LM' Released for Arabic LLMs
A new benchmark called 3LM has been released to evaluate the STEM and code generation capabilities of Arabic LLMs.
Gradio Supports Building Python-Based MCP Servers
Gradio's MCP integration feature automatically converts Python functions into LLM tools, enabling the construction of AI agents.
Hugging Face Unveils Trackio
Hugging Face has unveiled Trackio, a lightweight experiment tracking library that can be used as a drop-in replacement for wandb.
Hugging Face Introduces Parquet CDC Feature
Parquet CDC functionality has been added to PyArrow and Pandas, improving data transfer and storage efficiency on the Hugging Face Hub.
Hugging Face Unveils New CLI `hf`
Hugging Face has released a new CLI called `hf` with a redesigned command structure, along with `hf jobs`, a serverless job execution service.
How to Optimize Flux LoRA Inference Speed by 2.3x
A method using Diffusers and PEFT was released that improves LoRA inference speed for Flux models by about 2.3x.
Video LMM Benchmark 'TimeScope' Released
TimeScope, an open-source benchmark for measuring temporal understanding of long videos, has been released.
NVIDIA NIM Accelerates Hugging Face LLMs
NVIDIA NIM enables fast and stable deployment of over 100,000 LLMs from Hugging Face.
Arc unveils Virtual Cell Challenge
It introduces the Virtual Cell Challenge for developing AI models that predict gene silencing effects.
Gradio Significantly Enhances MCP Server Features
Gradio has improved file uploads, real-time progress notifications, OpenAPI integration, and authentication features for MCP servers.
FutureBench, a Benchmark for Predicting Future Events, Released
It proposes FutureBench, a new benchmark that evaluates AI agents' ability to predict future events.
Ettin: SoTA Encoder-Decoder Models
The Ettin Suite has been released, a family of SoTA-level encoder and decoder models trained with the same data and training recipe.
Hugging Face Switches Repository Backend to Xet
Hugging Face has switched the Hub repository backend from Git LFS to Xet to handle large-scale AI workloads.
Theorem-Proving SOTA Model Kimina-Prover Released
The Numina and Kimi teams released Kimina-Prover, which applies test-time RL search to achieve SOTA on the miniF2F benchmark.
Hugging Face Unveils Official MCP Server
Hugging Face has unveiled an official MCP server that enhances AI assistants' access to the Hub.
Docker-based Desktop Environment for GUI Agents, ScreenEnv, Released
It's a Python library that provides an isolated Ubuntu desktop environment within a Docker container for testing and deploying GUI agents.
Asynchronous Inference Technology That Solves Robot Control Latency
By using asynchronous inference that separates action prediction from execution, robot control latency is reduced and task speed is doubled.
Gradio Begins Supporting the MCP Protocol
With Gradio now supporting the MCP protocol, Hugging Face Spaces can be used as tools for LLMs.
AMD MI300X Optimized Kernels Released
Hugging Face and AMD collaborated to release custom kernels specifically for MI300X that maximize the performance of Llama 3.1 405B.
Hugging Face unveils open-source AI robot 'Reachy Mini'
Hugging Face and Pollen Robotics collaborated to launch 'Reachy Mini,' an affordable open-source AI robot.