EmbodiedLab Advanced Tier (T0 success)
Agents · 2026-10-26
The 12 advanced-tier EmbodiedLab experiments, where no fixed design meets the tolerance and the agent must choose the conditions, decide each next condition from earlier readings, identify a model or act on its result (a prediction or configuration whose physical continuation is scored). Same T0 interface, scoring gates and averaging as the full benchmark; the score is the success rate on these experiments, in percent.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 50.0 |
| GPT-6 Astra | 25.0 |
| GPT-6.1 Sol | 16.7 |
| DeepSeek V4.1 Flash 1017 | 0.0 |
| GLM-5.3 Flash | 0.0 |
| Kimi K3 | 0.0 |