Aim
Open-source, self-hosted experiment tracker with a fast UI for comparing many training runs; optional paid Team plan from AimStack
Overview
Aim is "an easy-to-use & supercharged open-source experiment tracker." You log runs from Python and open a local web UI built to compare many runs at once: metric explorers, grouping and aggregation, side-by-side views and a query language, plus a Python API for the same data. It integrates with common training libraries (Hugging Face and PyTorch Lightning among them), can import runs from other trackers and runs in Docker or on Kubernetes. Aim is Apache-2.0 licensed and maintained by AimStack, which also sells a Team plan. Activity has slowed: the latest tagged release is v3.29.1 from May 8, 2025, more than 18 months ago, although the repository still received pushes in November 2026 and is not archived.
The Verdict
Who Should Use Aim?
Best For
- Individuals and small teams that want a free, self-hosted tracker with a good comparison UI
- Projects with hundreds of runs to compare locally
- Teams that cannot send training metadata to a SaaS vendor
Not Ideal For
- Teams that need a model registry, deployment or an active release cadence
- LLM tracing and evals (MLflow, W&B Weave or Opik cover these)
- Organizations that want a large vendor behind their tracker
What's Great
- Free, Apache-2.0 and fully self-hosted
- Fast UI built for comparing many runs, with a query language
- Can migrate runs in from other trackers
- Low-cost paid Team plan if you want support or managed hosting
Watch Out For
- No tagged release since May 2025
- Fewer integrations than W&B
- Small company and community behind it
- No model registry or serving
Pricing
View all features & details
Key Features
- Run logging from Python
- Metric explorers, grouping and aggregation
- Side-by-side run comparison and query language
- Python query API
- Terminal output capture and notifications
Deployment
- pip install and local UI
- Docker image
- Kubernetes deployment guide
- Remote tracking server
Real-World Usage
Project Health
- Not archived; repository pushed in November 2026
- Last tagged releases: v3.28.0 (March 2025), v3.29.1 (May 2025)
Typical Users
- Researchers comparing many training runs locally
- Teams moving off a hosted tracker to a self-hosted one
How It Compares
| Feature | Aim | MLflow | Weights & Biases | ClearML |
|---|---|---|---|---|
| Model | Open source, self-hosted | Open source | Hosted SaaS | Open source plus SaaS |
| Model registry | No | Yes | Yes | Yes |
| LLM tracing | No | Yes | Yes (Weave) | GenAI App Engine |
| Paid entry | $11 per user per month | Managed via clouds | From $60 per month | $15 per user per month |
| Release cadence | Last release May 2025 | Frequent | Frequent | Frequent |
Sources: aimstack.io · Aim pricing · GitHub · mlflow.org
Is this your tool? Add a badge