Did your ChatGPT app ever flash a Thinking emoji and you wondered if it was doing anything beyond drama? That quirk — a toggle that looked like a deeper-thinking mode but simply routed replies through a standard setting — is being fixed. OpenAI has rolled out a genuine Thinking control in the ChatGPT Android app, and it arrives alongside a small but useful UI upgrade: formatting blocks that turn chat output into editable document-like pieces.
What changed on Android
If you use ChatGPT on Android and you’re a ChatGPT Plus subscriber, update the app and peek in chat settings. The Thinking control no longer plays dress-up. You can choose between Auto, Instant, and Thinking (Extended) modes.
- Instant favors speed and minimal inference — fast answers, lower compute.
- Auto picks a mode for you based on the prompt.
- Thinking gives the model more time — more compute and higher latency — to do multi-step reasoning and verification before replying.
- The Thinking toggle is currently available only to ChatGPT Plus subscribers; free users stay on the standard reasoning budget.
- iOS parity hasn’t arrived yet — iPhone users may still be waiting for the proper Thinking toggle.
- Expect a latency-versus-accuracy tradeoff: pick Instant for quick lookups, Thinking for careful work.
That mirrors the Extended Thinking option desktop users already had, but until now Android’s toggle was more placebo than power. The staged rollout means some users will see the switch immediately; others may need to wait for a server-side enable.
Why the extra thinking matters
Longer reasoning windows aren’t just vanity. On multi-step problems — math word problems, complex planning, multi-hop information synthesis, or careful code review — giving a model more time to internally check and compare options often yields fewer dumb mistakes and more complete responses. There’s a cost: more server compute and a slower answer. On mobile, though, having the choice is the point — triage when you need speed, lean in when you need depth.
This change also nudges ChatGPT on Android closer to the productivity workflows that competing systems are pushing on mobile. That arms race for better on-device AI experiences is one reason OpenAI has been iterating fast this year; you can see the same push for integrated productivity features in other platforms’ work on mail and documents, such as Google’s efforts to fold Gemini into Gmail and Drive Gemini Deep Research plugs into Gmail and Drive.
Formatting blocks: chat that feels like a mini editor
The other visible change is the introduction of formatting blocks — think of them as tiny document panes inside a chat. Ask ChatGPT to draft an email or a memo and the output appears in a block with a mini editor toolbar when you highlight text. Bold, italics, lists, and in-place edits are possible without asking the model to regenerate the whole response.
That’s a small UI tweak with outsized ergonomics. Instead of losing context to repeated regenerations, you can make line-level adjustments and nudge phrasing directly. For people using ChatGPT as a writing assistant or quick office tool, those blocks make the app act less like a chatroom and more like a lightweight text editor — a pivot towards productivity that OpenAI has been nudging at across features.
A few practical notes for users
If you follow OpenAI’s Android rollouts, this is consistent with the company putting more features on Google’s platform lately — from lab-style apps to broader releases — similar to other OpenAI moves like getting Sora on Android. The upgrades also reflect a larger trend: conversational AI increasingly behaves like a suite of productivity tools, not just a question-answer toy.
The change won’t suddenly make every answer perfect. But for users who regularly use ChatGPT for multi-step analysis, drafting constrained documents, or code review, the new toggle and inline editing can shave off frustration and reduce the number of follow-up prompts needed to get something usable.
If you don’t see the features yet, check for the latest update and be patient — these rollouts often arrive in waves. Meanwhile, expect more small, interface-first refinements: they’re quietly how companies are trying to make AI feel more useful in daily work.