IDEAL-Bench - Position Accuracy: leaderboard
Object position accuracy on IDEAL-Bench (2026), a single-image 3D indoor layout benchmark: the percent of matched objects in the seven non-grid room types whose predicted horizontal centre lies within 0.3 m of the ground truth. Higher is better.
Metric: Position accuracy (%): share of matched objects in the 7 non-grid room types whose horizontal centre error is at most 0.3 m, IDEAL-Bench 3D layout task; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o | 11.5 | #741 |
| 2 | Gemini 2.5 Pro | 10.8 | #266 |
| 3 | Gemma 4 31B (IT) | 9.8 | #342 |
| 4 | GPT-5.4 | 8.5 | #123 |
| 5 | Qwen 3 VL 8B Instruct | 8.2 | #981 |
| 6 | Qwen 3 VL 30B A3B Instruct | 7.5 | #871 |
| 7 | Qwen 3 VL 235B A22B Instruct | 7.2 | #522 |
| 8 | Claude Sonnet 4.6 | 4.9 | #161 |
| 9 | GLM-4.6V | 4.4 | #701 |
| 10 | Qwen 2.5 VL 72B Instruct | 2.1 | #846 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=ideal-bench-position-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-13.