distilabel
ArgillaFramework for building synthetic-data and AI-feedback pipelines: LLM-driven generation, labeling, and preference-data construction with pluggable model backends.
Argilla's original authors have moved on, and community collaborators now maintain the project.
Openness
5 high confidence- license
- Apache-2.0(OSI)
- source
- public(argilla-io/distilabel)
- core features withheld
- no
Fully OSI-licensed (Apache-2.0), full source public, no proprietary core.
- https://github.com/argilla-io/distilabel/blob/main/LICENSE recorded 2026-08-13
Apache License Version 2.0 text
- https://raw.githubusercontent.com/argilla-io/distilabel/main/LICENSE recorded 2026-09-29
LICENSE text reads "Apache License" and "Version 2.0, January 2004" followed by "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".
- https://raw.githubusercontent.com/argilla-io/distilabel/main/pyproject.toml recorded 2026-09-29
pyproject.toml describes "Distilabel is an AI Feedback (AIF) framework for building datasets with and for LLMs." and declares the license as "Apache-2.0". Searched the body for enterprise, license key and paid: none.
- https://raw.githubusercontent.com/argilla-io/distilabel/main/README.md recorded 2026-09-29
README says "A group of community members have recently joined the GitHub project as collaborators to maintain the project" and installs with "pip install distilabel --upgrade" from PyPI. Searched the body for enterprise, license key and paid: none.
Adoption
2 high confidencePyPI downloads for distilabel are the published usage signal, reflecting an established place in the Hugging Face synthetic-data workflow. Development responsibility has since passed from the original authors to community collaborators.
- https://pypistats.org/api/packages/distilabel/recent recorded 2026-08-13
16,308 downloads in the trailing 30 days for distilabel
Capability
4 medium confidenceIndirect proxy: tool quality inferred from models trained on data curated with the tool. Confidence capped because there is no head-to-head against other synthetic-data frameworks.
- https://huggingface.co/argilla/distilabeled-Hermes-2.5-Mistral-7B recorded 2026-09-29
Model card says "5,922</code> instead of <code>12,859</code> samples (54% reduction)" and reports Nous averages "54.04" and "53.51".
- https://huggingface.co/argilla/notus-7b-v1 recorded 2026-09-29
Model card table row for the Notus model gives AlpacaEval "91.42" and the Zephyr-7b-β row gives "90.60", with "we invested time understanding and deep-diving into the UltraFeedback dataset" in the text.
Verified 2026-09-24