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. Status: saturation imminent. 29 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Human Expert | 100 |
| 2 | Median Human | 90.7 |
| 3 | GPT-5.6 Sol (Max) | 66.7 |
| 4 | GPT-5.4 (High) | 50.6 |
| 5 | GPT-5.5 (High) | 46.1 |
| 6 | Qwen 3 VL 235B A22B Instruct | 39.4 |
| 7 | Claude Fable 5 | 35 |
| 8 | Gemini 3.1 Pro (Preview) | 32.2 |
| 9 | Gemini 3.5 Flash | 31.1 |
| 10 | Gemini 3 Pro | 28.9 |
| 11 | Grok 4.5 | 21.7 |
| 12 | Gemini 2.5 Pro | 18.9 |
| 13 | Claude Opus 4.7 | 15 |
| 14 | GPT-5.2 (High) | 15 |
| 15 | Qwen 3 VL 235B A22B (Thinking) | 14.4 |
Interactive version: theaggregate.ai/benchmark?slug=clockbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.