Embodied3DBench - Grasp Point (3D): leaderboard

Embodied3DBench (2026) Grasp Point (3D) task on simulated robotic tabletop scenes: grasp 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: Grasp Point (3D) task (grasp 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
1GPT-557#185
2Gemini 2.5 Flash53.5#466
3Gemini 2.5 Pro53.3#280
4Doubao-Seed-1.6Thinking52.2
5Seed-1.6Non-reasoning51.7#481
6Claude Sonnet 449#414
7Doubao-1.5-thinking-vision-pro47.4#368
8Qwen 3 VL 4B Instruct39.4#1348
9Qwen 2.5 VL 72B Instruct31.6#849
10GPT-4o29#754
11InternVL3-38B20.3#948
12InternVL3.5-8B19.6#1618
13Qwen 2.5 VL 7B Instruct12.6#1919

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=embodied3dbench-grasp-point-3d · How It Works · Data refreshed daily, snapshot 2026-11-22.