Ars Technica compared Google's Gemini 3.2 Fast and OpenAI's ChatGPT 5.2 using free-tier models. The test, prompted by Apple's decision to integrate Gemini into future Siri updates, aimed to reflect the experience for most users. Evaluations used new, more complex prompts, moving beyond the simple tests from 2023.

In the dad joke challenge, neither model excelled at originality. ChatGPT produced two seemingly original but awkward jokes, while Gemini's responses were easily found online. For a math problem estimating Windows 11's floppy disk size, both AI's calculations slightly underestimated the actual 6.7-7.2GB ISO size, showing similar reasoning flaws.

These results highlight stylistic differences between the two AI giants. While the test wasn't scientific, it reveals that for casual users, the gap between leading AI models is nuanced. The real impact may depend on specific tasks and integration, like Apple's pending Siri overhaul.