Argilla
Open-source tool for AI engineers and domain experts to build LLM datasets; part of Hugging Face, now in maintenance mode with no new features
Overview
Argilla is "a collaboration tool for AI engineers and domain experts to build high-quality datasets": a web UI where experts review, label and rank LLM outputs for fine-tuning, RLHF and evaluation. The team joined Hugging Face in June 2024, shipped Argilla 2.0 in July 2024 and added no-code dataset building on the Hugging Face Hub in 2.4 (November 2024). Status: the project is now in maintenance mode. The GitHub README says the original authors have moved on to other projects, no new features will be added, bug fixes and patches are still published as needed, and the maintainers are looking for contributors to take over. The last release is v2.8.0 (March 11, 2025), although the repository still saw pushes in November 2026. It is still usable: Apache-2.0, deployable in one click on Hugging Face Spaces or with Docker and Kubernetes. Its sibling project distilabel (synthetic data and AI feedback) carries the same "authors moved on" notice. We mark this page deprecated so new users pick an actively developed tool. Also from Hugging Face: the Hugging Face Hub, Transformers and TRL.
The Verdict
Who Should Use Argilla?
Best For
- Existing Argilla users who need a stable, free tool for LLM data review
- Small teams that want a quick annotation UI on Hugging Face Spaces
- Contributors willing to help maintain the project
Not Ideal For
- New projects that need ongoing feature development
- Video or complex image annotation
- Teams that need vendor support or an SLA
What's Great
- Free and open source (Apache-2.0), no SaaS lock-in
- One-click deploy on Hugging Face Spaces, plus Docker and Kubernetes
- Built for LLM data: expert feedback for fine-tuning, RLHF and evaluation
- Datasets can be built straight from the Hugging Face Hub without code
Watch Out For
- No new features: the original authors have moved on
- No release since March 2025
- Focused on text and LLM data; a poor fit for video or complex vision work
Pricing
View all features & details
Key Features
- Annotation UI where domain experts give feedback on AI outputs
- Datasets for fine-tuning, RLHF and evaluation
- No-code dataset creation from the Hugging Face Hub (2.4)
Deployment
- Hugging Face Spaces (recommended in the README)
- Docker
- Kubernetes
Real-World Usage
History
- June 13, 2024: Argilla announces it is joining Hugging Face
- July 30, 2024: Argilla 2.0 ships
- March 11, 2025: v2.8.0, the latest release
Community
- About 517 forks; open pull requests still waiting for maintainers
- Sibling project distilabel is community-maintained
How It Compares
| Feature | Argilla | Label Studio | Labelbox |
|---|---|---|---|
| 2026 status | Maintenance only | Active | Active |
| License | Apache-2.0 | Apache-2.0 | Commercial |
| LLM evaluation | Yes | Yes, plus agent traces | Yes |
| Hosting | Self-host or HF Spaces | Self-host or HumanSignal cloud | Labelbox cloud |
| Entry price | Free | Free self-hosted; $99/mo cloud | Free 500 LBU a month |
Sources: Argilla GitHub · labelstud.io · Labelbox docs
Is this your tool? Add a badge