How to Use Gemini Auto Models and Thinking Levels Free

Google changed the free Gemini app experience starting October 9, 2026. Users without a Google AI plan no longer pick models manually. The picker now shows only Auto with Gemini 3 models. At the same time, everyone gains new effort (thinking) levels: Low, Medium, and High.

These updates simplify the interface while giving control over reasoning depth. Free users still reach stronger models when needed, but most prompts stay on the fast Flash-Lite path. This guide shows exactly what changed, how Auto selection works, and how to use the new levels effectively.

What Changed for Free Gemini Users

Google’s support page “Changes to Gemini model access and limits” confirms the rollout. Starting October 9, 2026, users without an AI subscription see Auto as the only option in the model picker.

Google states: “Gemini will use auto model selection to send each prompt to the Gemini model that suits it best. Most prompts will go to Flash-Lite, our fastest model. Prompts that need deeper reasoning may go to Flash or Pro.”

If you turn off smart model selection, free chats stay on Flash-Lite for every response. Paid plans keep more manual choices, but Auto and the effort levels roll out broadly.

Previously, free users could switch among Flash-Lite, Flash, and Pro. The new system routes automatically. This reduces confusion for casual use while preserving quality on harder tasks.

How Auto Model Selection Works

Auto analyzes your prompt and routes it. Simple questions, summaries, or quick facts go to Flash-Lite. Tasks that need multi-step logic, coding help, or careful analysis can land on Flash or Pro.

You do not see the routing decision in the chat. Responses simply appear. The system prefers speed for most free-tier traffic so weekly compute limits last longer.

Smart model selection is on by default. Turning it off locks free users to Flash-Lite only. That option sits in settings if you want predictable, lightweight replies.

Understanding the New Thinking Levels

Effort levels replace the older Extended toggle. They match the thinking_level controls already available in Google AI Studio and Antigravity.

• Low: Quick and efficient. Best for short answers, lists, or simple rewrites. Uses fewer tokens and returns faster.
• Medium: Balanced depth. Handles most everyday tasks with solid reasoning without long delays.
• High: Extra thorough. Runs deeper analysis, checks steps, and produces more complete answers. Consumes more of your usage limit.

Higher effort improves thoroughness but counts more against your compute budget. Free users share the same three levels as subscribers. The difference is that Auto decides the underlying model while you control the effort.

These levels appear under the model selector once the update reaches your account. Rollout was already visible on most free accounts by the evening of October 9 (Pacific Time).

Step-by-Step: Using Auto and Effort Levels

1. Update the Gemini app to the latest version on Android or iOS. The change requires a recent build.
2. Open a new chat. Tap the model picker. You should see Auto (Gemini 3 models) as the only free choice.
3. Look for the effort or thinking level control. It usually sits near the picker or in the compose area.
4. Select Low for speed, Medium for normal work, or High when the answer needs careful reasoning.
5. Send your prompt. Auto routes it and the chosen effort level guides how deeply the model thinks.
6. Check usage if answers feel limited. Higher effort and complex prompts burn through the 5-hour and weekly compute limits faster.

You can switch effort levels mid-conversation. Auto re-evaluates each new prompt. If a reply feels too shallow, raise the effort and rephrase the request.

Tips for Better Results on the Free Plan

Keep most prompts on Medium. Reserve High for coding bugs, multi-step plans, or research summaries. Low works well for brainstorming short ideas or formatting text.

Break large tasks into smaller ones. Auto and lower effort handle pieces efficiently, then you can combine results. This stretches free limits further than one giant High-effort request.

If you need consistent Flash or Pro behavior, consider an AI Plus or AI Pro plan. AI Plus retains Flash access; AI Pro keeps Pro and adds Deep Think. Free Auto still covers the majority of daily use.

Developers who want full control can use the Gemini API with explicit thinking_level parameters. The app’s effort levels mirror those API options. See related developer coverage in our Android Studio Quail 4 Agent Mode guide.

When to Upgrade

Free Auto works for quick questions, drafts, and everyday research. You hit limits faster with media generation, Deep Research, or repeated High-effort sessions. The support page notes that premium models and extended thinking consume more of the compute allowance.

AI Pro users gain Deep Think (maximum parallel reasoning) that was previously Ultra-only. If your work involves frequent complex reasoning, the paid tiers remove routing surprises and raise overall limits.

Conclusion

The October 2026 Gemini update trades manual model choice for automatic routing and granular effort controls. Free users keep access to capable models when the prompt justifies it, while Flash-Lite handles the bulk of traffic. Low, Medium, and High levels let you decide how thoroughly Gemini reasons without leaving the free plan.

Update the app, experiment with the three effort settings, and adjust based on the task. For most people the new Auto experience feels simpler and still useful.

Sources

• Google Gemini Apps Help: Changes to Gemini model access and limits
• 9to5Google reporting on the October 9 rollout
• Google AI for Developers documentation on thinking levels