VIA-Bench (LLM Judge) - Gestalt Illusions: leaderboard
Metric: Accuracy (%) on the Gestalt illusions questions, answer letter read from the response by a GPT-4.1-mini judge; VIA-Bench's 1,004 human-verified multiple-choice questions on illusory and anomalous images, every option set including a never-correct 'Not Sure' option; zero-shot with an answer-format instruction, temperature 0.8 where supported (model defaults otherwise), mean of five runs; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 83.76 | #64 |
| 2 | Gemini 2.5 Pro | 65.81 | #145 |
| 3 | Claude Sonnet 4 (20250514) | 61.2 | #211 |
| 4 | Qwen 3 VL 235B A22B (Thinking) | 55.56 | #228 (Qwen 3 VL 235B A22B) |
| 5 | Claude Opus 4.1 (20250805) | 53.16 | #126 |
| 6 | GLM-4.5V | 49.57 | #339 |
| 7 | Qwen 2.5 VL 72B Instruct | 48.55 | #364 |
| 8 | Claude 3.5 Sonnet (20241022) | 46.84 | #286 |
| 9 | Qwen 3 VL 30B A3B (Thinking) | 45.47 | #338 (Qwen 3 VL 30B A3B) |
| 10 | GPT-5 Mini | 38.63 | #176 |
| 11 | O3 | 36.75 | #121 |
| 12 | Qwen 2.5 VL 7B Instruct | 34.19 | #643 |
| 13 | O4 Mini | 26.84 | #172 |
| 14 | GPT-5 Chat | 26.67 | #233 |
| 15 | GPT-4o (2024-11-20) | 22.39 | #369 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=via-bench-llm-judge-gestalt-illusions · How It Works · Data refreshed daily, snapshot 2026-10-11.