FREAK (Multiple Choice) - Consistency Accuracy: leaderboard
Metric: Consistency accuracy (%): share of FREAK's 1,000 multiple-choice questions answered correctly under all six cyclic permutations of the options; reasoning enabled where the model supports it; higher is better. Source: arxiv.org. Saturation forecast: Around 2032. 23 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3-78B | 35.9 | #345 |
| 2 | InternVL3-38B | 35 | #395 |
| 3 | GLM-4.5V | 34.3 | #339 |
| 4 | Gemini 2.5 Pro | 33 | #145 |
| 5 | O3 (High) | 32.4 | #121 (O3) |
| 6 | GPT-4.1 | 31.5 | #240 |
| 7 | Qwen 2.5 VL 72B Instruct | 30.4 | #364 |
| 8 | Qwen 2.5 VL 32B Instruct | 28.3 | #443 |
| 9 | O4 Mini | 27.6 | #172 |
| 10 | Qwen 2.5 VL 7B Instruct | 26 | #643 |
| 11 | InternVL3-14B | 25.3 | #494 |
| 12 | Gemini 2.5 Flash | 25.1 | #237 |
| 13 | InternVL3-8B | 24.8 | #606 |
| 14 | Phi-4 Multimodal Instruct | 19.2 | #896 |
| 15 | Claude Sonnet 4 (Thinking) | 18.5 | #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-consistency-accuracy · How It Works · Data refreshed daily, snapshot 2026-10-11.