How to Use Gemini Auto Model and Effort Levels

Google rolled out major changes to the Gemini app in October 2026. Free users without a Google AI plan can no longer pick models manually. The picker now shows only Auto. At the same time, everyone gains access to new effort levels that replace the old Extended thinking option.

These updates affect how you get answers. Auto routes most everyday prompts to Flash-Lite for speed. Harder tasks may go to Flash or Pro. Effort levels let you control how thoroughly Gemini reasons before responding. Higher effort gives deeper answers but consumes more of your usage limit.

This guide shows you exactly how Auto works, how to set effort levels, and practical ways to get better results on the free plan. You will also see what changes for paid tiers.

What Changed in the Gemini App for Free Users

Starting October 9, 2026, users without a Google AI plan see only Auto in the model picker. Google describes the behavior clearly on its support page: Gemini uses auto model selection to send each prompt to the model that suits it best. Most prompts go to Flash-Lite, the fastest model. Prompts that need deeper reasoning may go to Flash or Pro.

If you turn off smart model selection, free accounts use Flash-Lite for every response. You lose the chance that Auto will escalate a hard prompt.

Paid plans keep more choices. Google AI Plus users will soon be limited to Flash-Lite and the 3.8 Flash model. AI Pro and AI Ultra retain broader access, including Pro and Deep Think on higher tiers. Usage limits remain compute-based and refresh every five hours until the weekly cap.

How Auto Model Selection Works

Auto does not mean every answer is the weakest model. Google’s in-app text and support documentation state that most prompts route to Flash-Lite while prompts needing deeper reasoning can move to Flash or Pro. You cannot force Pro on a specific query as a free user. Auto decides.

This design keeps free usage predictable. Simple questions stay fast. Complex ones can still reach stronger models when the system judges them necessary. Turning smart model selection off removes that escalation path entirely.

For developers or power users who previously switched models mid-chat, the change requires a new habit: write clearer prompts so Auto can recognize when a task needs more capability.

Step-by-Step: Check Your Model Setting and Set Effort Levels

1. Open the Gemini app on Android or iOS, or go to gemini.google.com. Make sure the app is updated.
2. Start a new chat. Look at the model picker near the prompt box. Free users should see Auto listed.
3. Tap or click the model area to confirm the setting. You cannot switch away from Auto on a free account.
4. Locate the effort or thinking level control. It appears for available models. Choose Low, Medium, or High.
5. Send a test prompt. Observe response speed and depth. Switch the effort level and repeat the same prompt to compare.
6. Optional: Check Settings for smart model selection. Leave it on if you want Auto to escalate harder prompts.

Low effort prioritizes speed and efficiency. Medium balances depth and speed. High is extra thorough and uses more of your limit. These levels match the thinkingLevel controls available in Google AI Studio and Antigravity.

When to Use Each Effort Level

Use Low for quick lookups, short rewrites, simple summaries, or brainstorming lists. Responses arrive faster and consume less of your quota.

Choose Medium for most daily work: multi-step instructions, code snippets, email drafts, or moderate research. It gives enough reasoning without excessive delay.

Select High when the answer quality matters more than speed. Examples include complex problem solving, detailed analysis, long document summaries, or debugging tricky code. Higher effort improves thoroughness but reaches your limit sooner.

On free accounts, combine High effort with well-written prompts so Auto is more likely to route the request to a stronger model when needed.

Tips for Better Results on the Free Plan

Write specific prompts. Instead of “explain this,” say “explain the trade-offs of using Auto versus fixed models for a coding task and give three examples.” Clear intent helps Auto decide correctly.

Break large tasks into smaller ones. Free limits are compute-based. A series of focused prompts often stays within limits better than one massive request.

Test effort levels on the same prompt. You will quickly see which level gives the useful answer without unnecessary delay.

Update the app regularly. Google notes that the latest version delivers the best experience with these changes.

If you need reliable access to Pro or higher limits, consider a Google AI plan. AI Pro multiplies standard limits by four; AI Ultra goes higher. See Google’s support page for current plan details and how to manage subscriptions.

For related Gemini workflows, check our guide on analyzing data with Gemini and BigQuery.

What Paid Users Should Expect

Google AI Plus subscribers face upcoming limits to Flash-Lite and 3.8 Flash. AI Pro and Ultra keep more models and higher multipliers on usage limits. Effort levels roll out to all subscribers. Higher effort still consumes more compute regardless of plan.

Deep Think remains available on the highest tiers for maximum parallel reasoning. Most users will not need it daily.

Conclusion

Auto model selection and the new effort levels change how free Gemini users work. You no longer pick Flash or Pro yourself, but you gain finer control over reasoning depth. Most prompts stay fast on Flash-Lite. Complex ones can still reach stronger models when Auto judges them necessary.

Practice with the three effort levels. Write clearer prompts. Keep smart model selection enabled. These habits let you extract solid results from the free plan. If your workflow demands consistent Pro-level output or higher volume, the paid plans remain the direct path.

The changes rolled out quickly in early October 2026. Update your app and test a few prompts today to see how Auto and the effort controls behave on your account.

Sources

• Google Support: Changes to Gemini model access and limits (support.google.com/gemini/answer/17004136)
• 9to5Google reporting on the October 2026 rollout
• Google AI Studio and Gemini API thinking level documentation