Playable level environments

level_1187LIVE

Gameplay · EnvironmentsBAYER 8×8

Gameplay · Environments

3,200

playable levels you can plug agents into

Levels from real games packaged as headless environments with a step API, reward hooks and deterministic seeds, so agents can be trained and evaluated at scale.

Request a sample →

Specs

Levels: 3,200
Interface: Gym-style step API
Rendering: Headless or pixels
Determinism: Seeded
Delivery: Linux builds + Docker
License: Non-exclusive · training

Built for

01Reinforcement learning for game agents
02Generalization benchmarks across levels
03Self-play and curriculum research

Rights

Ownership verified
Store assets removed
No engine source
Secrets & player data scrubbed
Commercial training license

Sample record

level_1187/env.pyExample

env = traindata.make("level_1187", render="pixels", seed=7)
obs = env.reset()
for _ in range(1000):
    action = agent.act(obs)          # move, camera, jump, use
    obs, reward, done, info = env.step(action)
    if done:
        obs = env.reset()

For AI teams only

Request a sample pack

We review each request and send the Playable level environments sample pack and full statistics to your work email.

Licensing

How buying works

Evaluate the sample pack and statistics
Agree scope: full set or subset, exclusive or not
Sign a training-use license
Delivery via cloud storage with manifests and checksums

← Back to catalog

Figures shown are examples.