TriViewBench: leaderboard
Multimodal Large Language Models (MLLMs) demonstrate strong performance on standard visual question answering benchmarks, yet their scalability under controlled structural complexity remains poorly understood. We introduce TriViewBench, a controlled three-view visual reasoning benchmark constructed from synthetic 3D scenes with explicitly parameterized object count and occlusion. The benchmark con
Metric: Overall (self-reported). Source: benchmarklist.com. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 79.55 | #245 |
| 2 | Claude 3.7 Sonnet | 68.62 | #249 |
| 3 | Qwen 3 VL 32B Instruct | 61.38 | #283 |
| 4 | GPT-4o | 60.35 | #343 |
| 5 | Qwen 3 VL 8B Instruct | 60.32 | #420 |
| 6 | Qwen 3 VL 4B Instruct | 52 | #515 |
| 7 | Qwen 2.5 VL 32B Instruct | 44.59 | #453 |
| 8 | Qwen 2.5 VL 7B Instruct | 37.05 | #654 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=triviewbench · How It Works · Data refreshed daily, snapshot 2026-10-20.