Benchmarks / Reasoning
ClockBench
Uses the first-party analog-clock accuracy leaderboard. Score: overall accuracy. Collected by Jake Boggs' benchmark dashboard.
ClockBench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Opus 5.5 | max | 77.2% |
| 2 | GPT-6 Astra | max | 75.0% |
| 3 | GPT-6.1 Sol | max | 74.4% |
| 4 | GPT-5.6 Sol | max | 68.3% |
| 5 | Gemini 3.8 Flash | max | 66.7% |
| 6 | Opus 5 | max | 65.5% |
| 7 | Fable 5.1 | max | 59.4% |
| 8 | Gemini 3.7 Flash | max | 53.3% |
| 9 | GPT-5.4 | high | 50.6% |
| 10 | GPT-5.5 | high | 46.1% |
| 11 | Muse Spark 1.3 | max | 42.2% |
| 12 | Qwen 3 | high | 39.4% |
| 13 | Fable 5 | max | 37.2% |
| 14 | Gemini 3.5 Flash | xhigh | 34.4% |
| 15 | Gemini 3.1 Pro | high | 32.2% |
| 16 | Gemini 3 Pro | high | 31.1% |
| 17 | Grok 4.5 | max | 21.7% |
| 18 | Grok 4.6 | max | 20.0% |
| 19 | Gemini 2.5 Pro | high | 18.9% |
| 20 | GPT-5.2 | high | 15.6% |
| 21 | Opus 4.7 | max | 15.0% |
| 22 | o3 pro | high | 14.4% |
| 23 | o3 | high | 12.2% |
| 24 | Gemini 2.5 Flash | medium | 11.1% |
| 25 | GPT-5 | high | 11.1% |
| 26 | GPT-5 Pro | high | 11.1% |
| 27 | Mistral Medium 3.1 | none | 10.6% |
| 28 | Opus 4.6 | max | 10.0% |
| 29 | Opus 4.1 | high | 8.9% |
| 30 | GPT-5 Mini | high | 8.9% |
| 31 | Sonnet 4.5 | high | 7.8% |
| 32 | Sonnet 4 | high | 6.7% |
| 33 | GPT-4o | high | 6.7% |
| 34 | Qwen 2.5 | – | 6.1% |
| 35 | GPT-5 Nano | high | 3.9% |
| 36 | Grok 4 Fast | none | 3.3% |