Embodied-BenchForge - Success Rate: leaderboard
Task success percentage on all 220 Interactive Embodied Track tasks, with evaluated checkpoints and reasoning settings pinned to Appendix Table 21.
Metric: Success Rate (%, 220 interactive tasks). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 (High) | 77.73 | #51 (Claude Opus 4.7) |
| 2 | Gemini 3 Flash (Preview) (High) | 72.27 | #90 (Gemini 3 Flash (Preview)) |
| 3 | Qwen 3.6 35B A3B (Thinking) | 46.36 | #238 (Qwen 3.6 35B A3B) |
| 4 | Claude Sonnet 4.6 (High) | 16.82 | #94 (Claude Sonnet 4.6) |
| 5 | Claude Haiku 4.5 (Non-reasoning) | 5 | #287 (Claude Haiku 4.5) |
| 6 | Qwen 3.5 4B (Thinking) | 0 | #514 (Qwen 3.5 4B) |
| 7 | Qwen 3.5 2B (Non-reasoning) | 0 | #1114 (Qwen 3.5 2B) |
| 8 | Qwen 3.5 0.8B (Non-reasoning) | 0 | #1410 (Qwen 3.5 0.8B) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=embodied-benchforge-success-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.