VLM-DeflectionBench - Oracle Accuracy: leaderboard
Metric: Accuracy (%) of knowledge-based visual question answering in the Oracle retrieval scenario (gold evidence only), judged by GPT-4o with the SimpleQA protocol; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 73.1 |
| 2 | InternVL3-38B | 64.2 |
| 3 | Pixtral-12B | 62.7 |
| 4 | Qwen 2.5 VL 32B Instruct | 61 |
| 5 | Gemini 2.5 Pro | 59.8 |
| 6 | Gemma 3 27B | 59.5 |
| 7 | Gemini 2.5 Flash | 58.8 |
| 8 | GLM-4.1V-9B (Thinking) | 55.2 |
| 9 | Claude Opus 4 | 49.1 |
| 10 | Claude Sonnet 4 | 47.5 |
| 11 | Mistral Small 3.1 | 42.6 |
Interactive version: theaggregate.ai/benchmark?slug=vlm-deflectionbench-oracle-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-07.