Argilla is an open-source data annotation and curation tool for AI and LLM training data, acquired by Hugging Face in 2024. See features and use cases.
Argilla is an open-source data annotation and curation platform designed for building high-quality training and evaluation datasets for machine learning and large language models. It was created by a Madrid, Spain-based team founded by Daniel Vila Suero and Francisco Aranda, initially operating under the name Recognai before the Argilla product became the company's primary focus.
In 2024, Argilla was acquired by Hugging Face, one of the largest open-source AI and machine learning platforms, bringing Argilla's annotation technology and engineering team into the Hugging Face ecosystem. The team has continued active development post-acquisition, including the release of Argilla 2.0 and deeper Hugging Face Hub integration.
Argilla is used to collect structured human feedback on data for tasks ranging from text classification and named entity recognition to LLM preference tuning, RAG evaluation, and multimodal labeling.
Argilla provides a web-based review interface where domain experts and annotators can label, correct, and rank data, alongside a Python SDK for programmatic dataset management and integration into ML pipelines.
It supports human-in-the-loop workflows for a wide range of use cases including NLP classification, named entity recognition, RAG and retrieval evaluation, LLM preference and reward-model data collection, and multimodal (text-to-image) labeling.
Through Argilla Spaces on the Hugging Face Hub, teams can spin up hosted annotation environments and publish curated datasets directly as versioned Hugging Face datasets without managing their own infrastructure.
Argilla is free and open source; the core software can be self-hosted at no cost, with the full source code available on GitHub under an open-source license.
As part of Hugging Face, managed hosting options are available through Hugging Face Spaces and related infrastructure, generally following Hugging Face's own compute and hosting pricing rather than a separate Argilla-specific pricing plan.
Yes, Argilla is open source and free to self-host; managed hosting is available through Hugging Face Spaces.
Argilla was acquired by Hugging Face in 2024 and is now part of the Hugging Face ecosystem.
Argilla is used to annotate, label, and curate high-quality datasets for training and evaluating machine learning and LLM applications.
Argilla was founded by Daniel Vila Suero and Francisco Aranda in Madrid, Spain.
Yes, Argilla supports collecting human preference data, RAG evaluation, and other data used for LLM fine-tuning and evaluation.
Yes, Argilla integrates directly with the Hugging Face Hub, allowing curated datasets to be published and versioned there.
No, Argilla also supports multimodal annotation tasks such as text-to-image labeling.