Experimental: this provider has not been tested from this site yet. If the browser blocks the call, you will see why.
Pick a model, or choose Other model id to type one. If the provider does not know the model, the page offers the default to try.
Thinking: Minimal asks the model to think as little as it can (about 3 times faster in our test on one Anthropic model). Default leaves the provider's own setting, which can be slower and may cut a reply off on small models. This applies to Google and Anthropic models.
Detail sets how many boxes the question asks the model to aim for: Brief about 6, Normal up to 9, Detailed about 12. In our Bitcoin test Detailed drew about 12 boxes for roughly 0.16 s more model time. The model may not hit the number exactly; the ledger shows what the reply had.
Where your data goes
- Your question goes to Google (the Gemini API), under Google's terms, using your key. We never receive it.
- Free-tier prompts may be read by Google's reviewers. As of 16 November 2026, Google's Gemini API terms say that on the free tier "human reviewers may read, annotate, and process your API input and output", and that Google uses the content to improve its products. Paid use is treated differently. Check Google's current terms before you rely on any of this.
- The diagram text the model writes (not your key, not your question) goes to our drawing service. It draws the picture and keeps no copy. It writes one log line per request with the route, the result code, the size of the request and whether it was refused, and it keeps a count of drawings per day. It does not log the text, any key or the check token, and our code does not log your address. To limit abuse it counts requests under a one-way hash of your address and the date.
- Your key stays in this tab. It is held in memory, sent only to Google, never put in the address and never sent to us. A browser extension or a script on this page could read it, so use a throwaway key and delete it afterwards.