Compare

GPT-5.5 vs Gemini 3.1 Pro

One real example: the same prompt sent to both models, answers shown side by side exactly as they came back. It shows how each model handled this one task. It is not a benchmark and does not say which model is better in general.

Recorded on with GPT-5.5 (OpenAI) and Gemini 3.1 Pro (Google) through the Vercel AI Gateway, unedited. Same prompt, Converso's default system prompt, answers capped at 700 tokens. Reasoning for this recording: low (OpenAI, Google); off (Anthropic); in the app, Converso uses higher reasoning settings.

The prompt

Draft a short, friendly email to a client telling them their website delivery moves from Friday to next Tuesday because we are running a final round of testing. Keep it under 150 words.

GPT-5.5

OpenAI

Subject: Website Delivery Update

Hi [Client Name],

I wanted to let you know that we’re moving the website delivery from this Friday to next Tuesday.

We’re running one final round of testing to make sure everything is polished and working smoothly before handing it over to you. We appreciate your understanding and want to ensure you receive the best possible version.

Thanks again, and we’ll keep you updated if anything changes.

Best,
[Your Name]

gpt-5.594 output tokens

Gemini 3.1 Pro

Google

Subject: Quick update on your website delivery!

Hi [Client Name],

I hope you’re having a great week!

I’m reaching out with a quick update regarding your new website. To ensure everything is running flawlessly across all devices and browsers, we are conducting one final round of comprehensive testing.

Because of this, we are adjusting our delivery date from this Friday to next Tuesday, [Date]. We want to make absolutely sure your site is perfect before we hand over the keys, and this extra time will guarantee a smooth launch.

Please let me know if you have any questions. We can’t wait to share the finished product with you next week!

Best regards,

[Your Name] [Your Title/Company]

gemini-3.1-pro487 output tokens (331 reasoning)

Answers change from one prompt to the next, and from one run to the next. Judge on your own questions.

See more answers

The demo shows 3 prompts answered by three models side by side, including 2 more for this pair.

See the full comparison