CipherBank: leaderboard
Metric: Decryption Accuracy (%; 3-shot, mean of nine ciphers). Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.5 Sonnet (20241022) | 45.14 |
| 2 | O1 (2024-12-17) | 40.59 |
| 3 | DeepSeek R1 | 25.91 |
| 4 | O1 Mini (2024-09-12) | 20.07 |
| 5 | Gemini 2.0 Flash (Thinking) | 13.49 |
| 6 | DeepSeek V3 | 9.86 |
| 7 | Gemini 1.5 Pro | 9.54 |
| 8 | GPT-4o (2024-08-06) | 8.82 |
| 9 | Gemini 2.0 Flash (Preview) | 8.65 |
| 10 | GPT-4o (2024-11-20) | 6.4 |
| 11 | GPT-4o Mini (2024-07-18) | 1 |
| 12 | QwQ 32B-Preview | 0.76 |
| 13 | Qwen 2.5 72B Instruct | 0.55 |
| 14 | Llama 3.3 70B Instruct | 0.42 |
| 15 | Llama 3.1 70B Instruct | 0.38 |
Interactive version: theaggregate.ai/benchmark?slug=cipherbank · How It Works · Data refreshed daily, snapshot 2026-09-24.