Embodied3DBench - 2D Average: leaderboard
Metric: Mean of the five 2D task scores, each a normalized score (0-100): IoU of predicted boxes or the L2 error of predicted points (pixels in 2D, metres in 3D) mapped to 0-100, on the simulated robotic tabletop scenes of Embodied3DBench (21K QA pairs over 4.8K object instances), zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 4B Instruct | 56.5 |
| 2 | Qwen 2.5 VL 72B Instruct | 55.5 |
| 3 | GPT-5 | 54.4 |
| 4 | Qwen 2.5 VL 7B Instruct | 43 |
| 5 | Claude Sonnet 4 | 41.2 |
| 6 | GPT-4o | 36.6 |
| 7 | Gemini 2.5 Flash | 31.8 |
| 8 | Gemini 2.5 Pro | 31.4 |
| 9 | InternVL3-38B | 27.4 |
| 10 | Doubao-Seed-1.6 (Thinking) | 27.2 |
Interactive version: theaggregate.ai/benchmark?slug=embodied3dbench-2d-average · How It Works · Data refreshed daily, snapshot 2026-10-07.