AMA-Bench - Embodied AI: leaderboard
AMA-Bench embodied-AI slice measures memory and causal recall over embodied agent trajectories.
Metric: Average Score (%). Source: huggingface.co. Saturation forecast: Around July 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 56.44 | #110 |
| 2 | Qwen 3 32B | 48.19 | #443 |
| 3 | GPT-5 Mini | 41.31 | #186 |
| 4 | Gemini 2.5 Flash | 39.78 | #245 |
| 5 | Qwen 3 14B | 37.99 | #537 |
| 6 | Qwen2.5-14B-Instruct-1M | 33.16 | #578 |
| 7 | Claude 3.5 Haiku | 30 | #568 |
| 8 | Qwen 3 8B | 29.45 | #674 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ama-bench-embodied-ai · How It Works · Data refreshed daily, snapshot 2026-10-20.