FREAK (Multiple Choice): leaderboard
Metric: Accuracy (%) on FREAK's 1,000 multiple-choice questions, each asked under all six cyclic permutations of its three content options (option D fixed as none listed); reasoning enabled where the model supports it; higher is better. Source: arxiv.org. Saturation forecast: Around 2034. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 (High) | 43.98 | #121 (O3) |
| 2 | GPT-4.1 | 43.78 | #240 |
| 3 | InternVL3-78B | 43.63 | #345 |
| 4 | GLM-4.5V | 42.78 | #339 |
| 5 | Gemini 2.5 Pro | 42.73 | #145 |
| 6 | InternVL3-38B | 42.6 | #395 |
| 7 | O4 Mini | 41.58 | #172 |
| 8 | InternVL3-8B | 41.23 | #606 |
| 9 | Gemini 2.5 Flash | 41 | #237 |
| 10 | Phi-4 Multimodal Instruct | 37.56 | #896 |
| 11 | Qwen 2.5 VL 72B Instruct | 36.97 | #364 |
| 12 | Qwen 2.5 VL 7B Instruct | 36.08 | #643 |
| 13 | InternVL3-14B | 36.03 | #494 |
| 14 | Qwen 2.5 VL 32B Instruct | 36.03 | #443 |
| 15 | Claude Sonnet 4 (Thinking) | 30.3 | #194 (Claude Sonnet 4) |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=freak-multiple-choice · How It Works · Data refreshed daily, snapshot 2026-10-11.