OpenAI has been swinging for the fences with recent updates like GPT-5.2 and ChatGPT Images. This week, the company rolled out an improvement to its ChatGPT Android app, and it’s a change power users have been waiting for. ChatGPT’s official app now has a real Thinking mode, previously only available through a browser.
According to Bleeping Computer, the original Thinking toggle actually rerouted your responses through the Standard thinking time, which uses less compute power and doesn’t think for as long. The new toggle allows users to expressly choose between Auto, Instant, or Thinking modes to determine how long the AI will pontificate over a request before delivering an answer.
This matches the functionality already available for desktop users who choose “Extended Thinking,” which gives the AI more time to work on complex queries to deliver better, more accurate answers.
The rollout is happening now, so if you use ChatGPT on an Android device, make sure you’ve updated to the most recent version. Once done, the new Thinking mode will appear in your chat settings. From there, you’ll be able to choose between Standard and Extended thinking times, allowing you to get better answers on mobile. As of this writing, we have not been able to access the new thinking toggle in the ChatGPT iOS app.
This feature is only available for ChatGPT Plus users, so free users are still stuck with the old Thinking mode, which is really just Standard mode.
Per Bleeping Computer, the feature rolled out alongside a new formatting block feature. Now, when you ask ChatGPT to type an email, it does so in a formatting block, and you can request changes or edit the document without having to regenerate the entire email. This feature started rolling out over Christmas and is available to all users.
It’s a small but pleasant end to the year for ChatGPT users. OpenAI spent the latter half of 2025 allegedly in a “code red” state after losing a sizable number of users to Google’s Gemini 3 and Nano Banana, and has spent much of the rest of the year trying to recapture those people with GPT-5.2 and ChatGPT Images. It has mostly worked, as GPT-5.2 trades blows with Gemini 3 on most benchmark scores while soundly beating Grok 4 in most areas.
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
