Gym-Anything-RT (real-time minigames, paused)
Games · 2026-10-27
The same 100 Gym-Anything-RT real-time minigames at difficulty 5 with full interaction, run in paused execution: each observation advances the task through a short observation window (80-800 ms, three to six frames) and the clock then stays frozen until the model's action executes, removing the need to respond in time. One attempt per task; the score is the share solved, in percent.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 75.0 |