80%
#1 of 7
CalendarQA accuracy of GPT-o1, the best of seven models; Claude-3.5-sonnet scored 46.7%
What the chart shows
In the same study as the clock test, Saxena et al. (2025) asked seven multimodal models questions about yearly calendar images. GPT-o1 answered 80% correctly, well ahead of Claude-3.5-sonnet (46.7%) and GPT-4o (43.3%).
Every model did better on calendars than on clocks: Gemini-2.0, best at reading clocks with 22.6%, scored 31.7% here. The report notes more errors when questions needed date arithmetic than when they asked about well-known holidays.
Key numbers
| Measure | Value |
|---|---|
| GPT-o1 | 80% |
| Claude-3.5-sonnet | 46.67% |
| GPT-4o | 43.33% |
| Gemini-2.0 | 31.67% |
| Llama 3.2-Vision | 11.67% |
How to read it
- Accuracy is the share of date questions answered correctly.
- Every value is a multiple of 1/60 (about 1.67%), consistent with a small set of 60 questions, so differences of a few points are not meaningful.
- Clock-reading results for the same models are in the related chart.
Source: Saxena et al., 2025, via the AI Index Report 2026 (Stanford HAI), figure 2.4.7: CalendarQA: accuracy. Chart and text by AI Adoption.
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Numbers checked against the published data on 20 October 2026.
Download the data (CSV) · Cite this chart
Discussion