Embodied3DBench - 3D Average: leaderboard

Embodied3DBench (2026) low-level embodied spatial intelligence in metric 3D: a vision-language model predicts 3D boxes, points, spatial relations, affordances, trajectories and grasp points in camera space on simulated robotic tabletop scenes. Mean of six normalized 0-100 task scores derived from IoU and distance in metres; higher is better.

Metric: Mean of the six 3D 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: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-547.8#178
2Gemini 2.5 Pro44.1#276
3Gemini 2.5 Flash41.6#463
4Seed-1.6Non-reasoning41.1#469
5Doubao-Seed-1.6Thinking40.9
6Qwen 3 VL 4B Instruct40.1#1332
7Claude Sonnet 439.7#410
8Doubao-1.5-thinking-vision-pro38.3#344
9GPT-4o31.3#751
10InternVL3-38B28.2#939
11Qwen 2.5 VL 72B Instruct27.5#841
12InternVL3.5-8B25.2#1604
13Qwen 2.5 VL 7B Instruct23.3#1902

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=embodied3dbench-3d-average · How It Works · Data refreshed daily, snapshot 2026-11-21.