SIMA
Instructable agent that follows natural-language commands across nine 3D games using only screen pixels and keyboard and mouse.
Trained with eight game studios on nine games plus a Unity research environment. Evaluated on about 600 basic skills across 1,500 unique tasks by human judges; multi-game training beat single-game specialists.
- Date
- Wednesday, 13 March 2024
- Lab
- Google DeepMind
- Kind
- model
- Access
- research preview
Figures
| Measure | Value | Measured by |
|---|---|---|
| Evaluated tasks | ~1,500 across ~600 skills human-judged | company |
Successor SIMA 2 announced 2025-11.
Sources
This record was checked against its sources on 6 October 2026. How we check