Gemini, Extended mode, better recall
Gemini integration
After seeing how much Gemini dominated in our our new LLM proofreading benchmark, we've added support for Gemini 3 Flash!
In initial testing, Gemini 3 Flash is the fastest model in Revise and will use less of your limit than other flagship models like GPT 5.4 or Claude Sonnet. We have noticed Gemini can be a bit more eager than other models and make changes beyond the scope of the initial request, which is something we will keep an eye on as the integration settles.
Extended mode update
Extended Mode is now a simple on/off toggle in the AI chat interface. Click the new button to turn it on.
When in Extended Mode, the agent will work for much longer before pausing to check in with you. This is useful for getting large scale edits done, but use it mindfully because you can also go through your monthly limit faster with vague/open-ended prompts.
Extended Mode is available under the Revise Plus and Pro plans.
Bug fixes
- The Agent now has better continuity around pending proposed changes and selected text.
- Fixed a bug that could cause text to get scrambled after an agent edit was made on formatted text.
- Fixed a case where the agent could incorrectly act like it did not have access to the document.
This should make the chat workflow feel more predictable, especially when you are revising large documents over multiple turns and looking at diffs while you talk to the agent.