KOFFVQA - Hallucination and Robustness: leaderboard
KOFFVQA subscore for robust visual answering while avoiding hallucinated details.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 82 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 Mini | 90 | #355 |
| 2 | Llama 4 Scout Instruct | 90 | #539 |
| 3 | Gemini 2.5 Flash (Preview 04-17) | 90 | #237 |
| 4 | InternVL3-38B | 86 | #405 |
| 5 | Qwen 2.5 VL 32B Instruct | 86 | #453 |
| 6 | InternVL2.5-78B | 85 | #431 |
| 7 | GPT-4.1 | 80 | #248 |
| 8 | Claude 3.5 Sonnet (20241022) | 80 | #291 |
| 9 | InternVL3-78B | 80 | #352 |
| 10 | GPT-4 Turbo | 80 | #404 |
| 11 | O4 Mini (2025-04-16) | 80 | #179 |
| 12 | O3 (2025-04-16) | 80 | #121 |
| 13 | Qwen 2.5 VL 72B Instruct | 80 | #377 |
| 14 | O1 (2024-12-17) | 80 | #149 |
| 15 | Mistral Small 3.1 | 80 | #610 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=koffvqa-hallucination-and-robustness · How It Works · Data refreshed daily, snapshot 2026-10-20.