How to Use Gemini Auto and Thinking Levels on Android
Step-by-step guide to Gemini Auto model selection and Low/Medium/High thinking levels for free and paid Android users in 2026.
Google updated the Gemini app so free users on Android now default to Auto model selection. Most prompts route to Flash-Lite for speed, while complex ones can go to stronger models. Everyone also gains Low, Medium, and High thinking levels that replace the old Extended option.
This guide walks through the exact changes, how to control effort, and practical ways to get better results without a paid plan. Facts come from Google’s rollout notes and 9to5Google reporting on the October 2026 update.
What Changed for Free Gemini Users
Free accounts no longer show a manual model picker. The only choice is Auto with Gemini 3 models. Google’s description states that Auto sends each prompt to the model that fits best. Most everyday questions land on Flash-Lite. Prompts that need deeper reasoning can shift to Flash or Pro.
If you turn off smart model selection, every response stays on Flash-Lite. Paid Google AI Plus users will soon be limited to Flash-Lite and Flash as well. Pro and Ultra plans keep broader access.
These limits rolled out rapidly starting October 9, 2026. Most free accounts already see the new picker.
Understanding Thinking Levels
Thinking levels (also called effort levels) let you control how thoroughly Gemini reasons. They align with the same controls in Google AI Studio and Antigravity.
- Low: Quick and efficient. Best for short answers, simple lookups, and fast drafts.
- Medium: Balanced depth. Good default for most tasks that need a little more structure.
- High: Extra thorough. Use for multi-step problems, coding, research summaries, or when accuracy matters more than speed.
Higher effort consumes more of your usage quota. Free users still hit daily limits, so reserve High for important prompts.
How to Access Auto and Thinking Levels
- Open the Gemini app on your Android phone or tablet.
- Start a new chat or open an existing one.
- Tap the model name or selector near the prompt box (it now shows Auto for free users).
- Confirm Auto is selected. You cannot switch to a specific model on a free plan.
- Look for the thinking or effort control. It appears as Low / Medium / High options under or beside the model selector.
- Choose the level before you send your prompt. You can change it mid-conversation for the next message.
On some builds the control sits in the prompt bar overflow menu. Update the Gemini app through the Play Store if you do not see it yet.
Practical Examples for Everyday Use
For a quick weather check or reminder: set Low. Auto routes it to Flash-Lite and returns an answer in seconds.
For summarizing an email thread or rewriting a message: try Medium. The response stays readable without extra fluff.
For debugging a snippet of code, planning a multi-stop trip, or comparing product specs: switch to High. Auto may escalate the model, and the higher effort gives more complete reasoning.
You can also attach screen context or files as usual. Thinking level applies on top of those inputs.
Tips to Stay Within Free Limits
Keep most prompts on Low or Medium. Save High for tasks where a short answer fails.
Turn smart model selection back on if you disabled it. Otherwise you stay locked to Flash-Lite even for hard questions.
Clear long conversation history when you notice slower or shallower replies. Context length still affects quality.
Combine with other free Android tools. For example, use Gemini’s screen context for quick app help, then switch to High only when you need a full walkthrough. See our earlier guide on Gemini Intelligence features for related agent-style workflows.
When to Consider a Paid Plan
If you regularly need Pro-level reasoning or hit free quotas, Google AI Pro restores fuller model choice and higher limits. AI Ultra adds Deep Think options on supported models. Check the official Gemini help page for current pricing and regional availability, as plans change.
Free Auto still covers the majority of daily Android tasks—timers, quick research, message drafts, and calendar checks.
Conclusion
Auto model selection and the three thinking levels give free Gemini users clearer controls without the old model dropdown. Start with Medium for most chats, drop to Low for speed, and raise to High only when the answer needs extra depth. The change keeps Gemini useful on Android while Google manages compute.
Update your app, test the new selector, and adjust effort based on the task. You will get more consistent results inside the free tier.
Sources
- 9to5Google: Free Gemini users now on ‘Auto’ models as app adds thinking levels (October 2026)
- Google Gemini support documentation referenced in the rollout (model selection and Auto behavior)
- Official Android and Gemini app updates confirming thinking level alignment with AI Studio
Share this story
