ClockBench: leaderboard
Visual reasoning benchmark testing if AI can read analog clocks. 180 clocks, 720 questions covering time reading, arithmetic, rotation, and timezone conversion. Human baseline: 90.7%.
Metric: Accuracy (%). Source: clockbench.ai. Saturation forecast: Around January 2027. 41 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Median Human | 90.7 | |
| 2 | Gemini 4 Argon (High) | 83.3 | #8 (Gemini 4 Argon) |
| 3 | Claude Opus 5.5 (Max) | 77.2 | #3 (Claude Opus 5.5) |
| 4 | GPT-6 Astra (Max) | 75 | #6 (GPT-6 Astra) |
| 5 | GPT-6.1 Sol (Max) | 74.4 | #7 (GPT-6.1 Sol) |
| 6 | GPT-5.6 Sol (Max) | 68.3 | #19 (GPT-5.6 Sol) |
| 7 | Gemini 3.8 Flash | 66.7 | #21 |
| 8 | Claude Opus 5 (Max) | 65.5 | #12 (Claude Opus 5) |
| 9 | Claude Fable 5.1 (Max) | 59.4 | #9 (Claude Fable 5.1) |
| 10 | Gemini 3.7 Flash | 53.3 | #23 |
| 11 | GPT-5.4 (High) | 50.6 | #79 (GPT-5.4) |
| 12 | GPT-5.5 (High) | 46.1 | #31 (GPT-5.5) |
| 13 | Muse Spark 1.3 | 42.2 | #24 |
| 14 | Qwen 3 VL 235B A22B Instruct | 39.4 | #273 |
| 15 | Claude Fable 5 | 37.2 | #16 |
No result here: #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=clockbench · How It Works · Data refreshed daily, snapshot 2026-10-20.