You Can Choose New 'Thinking Levels' For Gemini

Why It Matters

By letting users balance speed against depth, Google enhances Gemini’s utility for a broader audience and signals a push toward more granular AI control ahead of major product announcements.

Key Takeaways

• Gemini app adds "Thinking level" selector for Fast and Pro models
• Two modes: Standard (default) and Extended (more processing time)
• Extended mode improves answers on complex queries without using full Thinking model
• Feature rolls out to free users, not limited to AI Plus subscribers
• Launch precedes Google I/O, hinting more Gemini enhancements

Pulse Analysis

Google’s Gemini platform has long differentiated between "thinking" models that chain multiple reasoning steps and faster, single‑pass variants. The latest update introduces a "Thinking level" option directly in the mobile app, letting users decide how much computational effort the model should expend on a given prompt. While the Fast (Gemini 3 Flash) and Gemini 3.1 Pro models remain non‑thinking by default, the new Standard and Extended settings provide a middle ground: Standard for routine queries and Extended for topics that benefit from deeper analysis without invoking the full‑blown Thinking model.

For professionals who rely on AI for drafting reports, coding assistance, or data synthesis, this granular control can translate into measurable productivity gains. Extended mode adds latency—often a few extra seconds—but it can reduce the need for follow‑up prompts, delivering more comprehensive answers on the first try. The rollout to free users democratizes this capability, narrowing the feature gap between casual and AI‑Plus subscribers and potentially raising user expectations for responsiveness across competing platforms such as OpenAI’s ChatGPT and Anthropic’s Claude.

The timing is strategic. Unveiling the toggle just before Google I/O suggests the company will showcase additional Gemini enhancements, possibly integrating multimodal inputs or tighter integration with Android services. As enterprises evaluate AI partners, the ability to fine‑tune performance versus cost becomes a differentiator. Google’s move signals a broader industry trend toward customizable AI experiences, where developers and end‑users alike can tailor model behavior to fit specific workflow demands.