Model Hubs & Serving

Retro

Play classic games in Gym for reinforcement learning.

Status
UnmaintainedGitHub repository has been archived.
Last checked
Aging · Sep 3, 2026
License
Open source
Pricing
Free tier · from $0
Visit Retro Compare with alternatives

What it is

Retro provides a collection of classic video game environments for use with the OpenAI Gym framework, enabling developers to train and test reinforcement learning algorithms on these games.

Retro stands out as a specialized tool within the OpenAI Gym ecosystem, offering a unique collection of classic game environments tailored specifically for reinforcement learning research and education.

When it fits — and when it doesn’t

A good fit for

  • Researchers looking to benchmark their reinforcement learning models against classic game environments.

  • Students and educators who want to teach reinforcement learning using familiar video games.

Look elsewhere if

  • Developers needing real-time interaction with the environment for applications like robotics or autonomous vehicles.

  • Teams working on large-scale distributed training systems that require cloud-based solutions.

Known weaknesses

Each one links to the public issue, changelog or docs page it came from, where we have one.

  1. 01

    Limited support for modern games

    Retro primarily supports classic video game environments which may not be suitable for developing advanced reinforcement learning algorithms.

  2. 02

    Performance issues with complex game states

    Games with high-resolution graphics or complex state spaces can lead to slow performance and increased computational requirements.

  3. 03

    Documentation lacks depth for advanced use cases

    The documentation focuses on basic setup and usage, but does not provide detailed guidance for more sophisticated reinforcement learning tasks or custom game environments.

What it does well

  • Integration with OpenAI Gym for reinforcement learning.

  • Supports a wide range of classic video games.

  • Allows developers to train AI agents on game environments.

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 Retro

Setup steps, common gotchas and what to check after 30 days.

Open the setup guide →