Ultimate3D - Orientation: leaderboard
Off-the-shelf MLLM 8-way 3D object-orientation recognition accuracy (%) across combined synthetic and real images in Ultimate3D (arXiv:2611.24795, Table 1).
Metric: Orientation Accuracy (%), Synthetic + Real. Source: arxiv.org. Saturation forecast: Around September 2027. 20 models tracked. Updated 20 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3.5 Flash | High | 65.3 | #108 |
| 2 | GPT-5.5 | 54.1 | #54 | |
| 3 | Gemma 4 26B A4B (IT) | 47.3 | #561 | |
| 4 | Claude Opus 4.7 | 46 | #85 | |
| 5 | Gemma 4 31B (IT) | 44.9 | #354 | |
| 6 | GPT-4o Mini | 43.5 | #1678 | |
| 7 | GPT-4o | 43.1 | #755 | |
| 8 | Claude 3 Sonnet | 43.1 | #2338 | |
| 9 | Claude 3.5 Sonnet | 40.3 | #761 | |
| 10 | Qwen 3.5 9B | 39.6 | #829 | |
| 11 | Qwen 3.5 4B | 37 | #1235 | |
| 12 | Claude Sonnet 4.6 | 36.4 | #175 | |
| 13 | gemma-4-E4B-it | 29.8 | #1535 | |
| 14 | Qwen 3.5 2B | 28 | #5274 | |
| 15 | Gemma 4 E2B (IT) | 25.6 | #2449 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=ultimate3d-orientation · How It Works · Data refreshed daily, snapshot 2026-11-20.