Hugging Face

An efficient French medical encoder with long-context support.

A French medical encoder based on RoBERTa architecture.

LLM-rephrased subset of FineMed for enhanced AI training.

A French medical pretraining corpus for AI models.

Enhances GRPO with efficient, in-process AI generation capabilities.

New benchmarks for evaluating PEFT techniques in AI models.

Reference implementation of ARD for AI resource discovery.

Prevents reward hacking in GLM-5.2's RL training.

Supports training and inference for large-scale agentic RL.

Balances model capability against speed and computational cost.

Reduces computational cost in GLM-5.2's sparse attention layers.

Flagship model for long-horizon tasks with 1M-token context.

AI-driven game using LLM for storytelling and humor.

LoRA adapter for Whisper to clean disfluent speech transcription.

Enhanced leaderboard with improved speed, filtering, and transparency.

Interoperability layer for RL environments with standardized interfaces.

Private storage product optimized for AI artifacts.

Provides agents with a compact reference of commands.

Optimized CLI for coding agents using Hugging Face Hub.

Integrates remote tools into Reachy Mini via Hugging Face Spaces.

Native support for function-calling protocols in agent stacks.

Expanded capabilities for mobile environments, improving performance.

Enhanced computer-use model for diverse environments and frameworks.

Optimized quantized checkpoints for fast local inference.

Sparse delta encoding for efficient AI model weight updates.