TriViewBench - Level 4: leaderboard
TriViewBench (2026) tests multi-view structural reasoning in multimodal models on synthetic 3D tower scenes shown as front, side and top views, with controlled object count and occlusion; models answer directly within 128 tokens. This board is accuracy in percent on the 4,231 Level 4 questions (many objects, high occlusion); higher is better.
Metric: Accuracy (%) over the 4,231 Level-4 questions (high object count, high occlusion) of all three reasoning categories; three orthographic views (front, side, top) of synthetic 3D tower scenes, direct prompting with a concise answer within 128 tokens, greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 65.94 | #453 |
| 2 | Claude 3.7 Sonnet | 54.9 | #458 |
| 3 | Qwen 3 VL 8B Instruct | 46.68 | #981 |
| 4 | Qwen 3 VL 32B Instruct | 43.02 | #585 |
| 5 | GPT-4o | 42.76 | #741 |
| 6 | Qwen 3 VL 4B Instruct | 40.75 | #1279 |
| 7 | Qwen 2.5 VL 32B Instruct | 32.1 | #1053 |
| 8 | Qwen 2.5 VL 7B Instruct | 25.05 | #1833 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=triviewbench-level-4 · How It Works · Data refreshed daily, snapshot 2026-11-13.