Embodied3DBench - Manipulation Trajectory (3D): leaderboard

Embodied3DBench (2026) Manipulation Trajectory (3D) task on simulated robotic tabletop scenes: end-effector key trajectory points in metric 3D, predicted zero-shot by a vision-language model. Normalized 0-100 score derived from box IoU or point distance; higher is better.

Metric: Manipulation Trajectory (3D) task (end-effector key trajectory points in metric 3D), 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
1Qwen 3 VL 4B Instruct44.5#1332
2Gemini 2.5 Pro43.4#276
3GPT-538#178
4Gemini 2.5 Flash36.9#463
5Doubao-1.5-thinking-vision-pro34.7#344
6Doubao-Seed-1.6Thinking33.8
7Claude Sonnet 432#410
8Seed-1.6Non-reasoning29.9#469
9InternVL3-38B13.2#939
10GPT-4o11.8#751
11InternVL3.5-8B6.7#1604
12Qwen 2.5 VL 7B Instruct5#1902
13Qwen 2.5 VL 72B Instruct3.8#841

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

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