# openai/retro

Retro Games in Gym

Repository: https://github.com/openai/retro
Canonical: https://ross.abutalabs.com/products/retro
Language: C
License: MIT
License Family: permissive
Archived: true
Last push: 2024-02-22T13:04:14+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 3130, "days_push": 923, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3585, forks 535 (observed 2026-08-28T04:08:11.100850+00:00)

## What it is
Gym Retro is a Python library that turns classic video games into Gym environments for reinforcement learning research, with integrations for ~1000 games via Libretro emulators. It includes memory locations, reward functions, and savestates for each game integration, though ROMs must be supplied by the user.

## Use cases
- train reinforcement learning agents on retro video games
- benchmark RL generalization across game levels
- create Gym environments from emulated games
- research reward functions from in-game memory variables
- run RL experiments on Atari, NES, SNES, Genesis, and Game Boy games

## When to choose
- you need standardized RL environments based on classic games
- you want reproducible game integrations with savestates and reward definitions
- you need a benchmark for generalization in reinforcement learning

## When to avoid
- you need actively developed features or support for recent Python versions
- you want a general-purpose game emulator for playing games
- you expect ROMs to be bundled with the library

## Facets
- artifact type: library
- maturity: maintenance
- function: reinforcement-learning, simulation, game-engine, sdk
- domain: reinforcement-learning, machine-learning, artificial-intelligence
- platform: python, windows, cpp, c
- tags: gym-environments, emulation, libretro, retro-games, rl-benchmark, game-development, linux, macos

## Member repositories
- openai/retro (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:11.100850+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:33:55.406906+00:00, confidence not recorded.
  - readme: https://github.com/openai/retro (fetched 2026-08-28T04:08:11.100850+00:00, sha 000ca62aa98d)
- Data as of 2026-08-30T08:39:29.467469+00:00.