MetaWorld
Open-source robotics benchmark for meta- and multi-task reinforcement learning
- Status
- Generally available
- Last checked
- Checked · Oct 12, 2026
- License
- Open source
- Pricing
- Free tier · from $0
What it is
MetaWorld is an open-source framework designed to facilitate research in meta-learning and multi-task reinforcement learning, providing a suite of tasks that simulate real-world robotic scenarios.
MetaWorld stands out by offering a diverse set of robotic tasks specifically tailored for meta- and multi-task reinforcement learning, making it an essential benchmarking suite in the robotics research community.
When it fits — and when it doesn’t
A good fit for
Academic researchers looking to benchmark their multi-task reinforcement learning models against a standardized set of tasks.
Engineers developing autonomous robotic systems who need a comprehensive suite of simulated environments for testing.
Look elsewhere if
Developers seeking real-time robotics solutions as MetaWorld is primarily a research tool.
Teams requiring cloud-based services, as it is designed for local execution.
Known weaknesses
Each one links to the public issue, changelog or docs page it came from, where we have one.
- 01
Limited support for non-robotic scenarios
Framework is heavily optimized for robotic tasks, making it less suitable for other domains of reinforcement learning
- 02
Performance issues with large-scale simulations
Simulations involving a high number of parallel tasks can lead to significant performance degradation and increased computational requirements
What it does well
Suite of diverse robotic tasks for benchmarking
Supports meta-learning and multi-task learning
Extensive documentation and community support
Pricing
- Free tier
- YesOpen source — free to use
- Paid from
- $0
- Model
- Flat rate
- Enterprise plan
- No
Pricing not independently confirmed — check the vendor’s page.
Performance benchmarks
How Fast Is It?
Set up MetaWorld
Setup steps, common gotchas and what to check after 30 days.