Models & Datasets
Open-source foundation models and high-quality datasets for the community.
A highly efficient 7B parameter model optimized for reasoning and coding tasks. Outperforms Llama-2-13B on standard benchmarks.
State-of-the-art vision-language model capable of detailed image captioning, visual QA, and object detection.
High-fidelity text-to-audio generation model supporting sound effects, music, and speech in 40+ languages.
A curated dataset of 5M high-quality instruction-following pairs, filtered for safety and educational value.
Specialized coding model trained on 1T tokens of code across 50+ programming languages. Supports FIM and long context.
Comprehensive safety evaluation suite for testing LLM robustness against jailbreaks, bias, and toxicity.
Explore our full library on Hugging Face
Access all our models, datasets, and demos directly on the Hugging Face Hub. Join the community discussion and contribute.