Ego2World: leaderboard
Embodied agents in household environments must plan under partial observation: they need to remember objects, track state changes, and recover when actions fail. Existing benchmarks only partially test this ability. Egocentric video datasets capture realistic human activities but remain passive, while interactive simulators support execution but rely on synthetic scenes and hand-crafted dynamics,
Metric: Goal Tasks (self-reported). Source: benchmarklist.com. Saturation forecast: Around September 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5 | 183 | #141 |
| 2 | Gemini 3.1 Pro (Preview) | 171 | #58 |
| 3 | Claude Sonnet 4.6 | 171 | #90 |
| 4 | GPT-5.5 | 171 | #31 |
| 5 | Qwen Plus | 140 | #385 |
| 6 | DeepSeek V4 Flash | 117 | #137 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ego2world · How It Works · Data refreshed daily, snapshot 2026-10-20.