Google has reportedly begun rolling out a new set of features across its Phone and Gemini apps, introducing upgrades that focus on clearer communication and more advanced artificial intelligence reasoning for power users.
Expressive Calling starts reaching beta users
Following an announcement earlier this month, Google’s new Expressive Calling feature is now appearing for some users enrolled in the Phone by Google beta programme, reported 9To5Google. The feature is designed to make calls feel more urgent and noticeable through visual cues and haptic feedback.
Users can check availability by heading to Phone Settings, selecting General and scrolling down to Expressive Calling. The main toggle, which enhances calls with visual and vibration effects, is switched on by default. There is also an optional setting that allows urgent calls to break through Do Not Disturb mode.
How urgent calls work
At this early stage, Expressive Calling appears to work only when both the caller and the recipient are using the beta version of Phone by Google, reportedly version 203. When eligible, callers see a prompt asking whether they want to mark the call as urgent.
If selected, the recipient is shown a clear “It’s urgent!” message on the incoming call screen, accompanied by an animated siren emoji. Should the call go unanswered, it is later marked as urgent in the call history, making it easier to identify missed important calls.
Gemini 3 Deep Think unveiled for AI Ultra users
Meanwhile, Google officially introduced Gemini 3 Deep Think earlier this month, within the Gemini app. This new reasoning mode is available to subscribers on the Google AI Ultra plan and is positioned as the company’s most advanced AI thinking capability to date.
The rollout is aimed at users who regularly deal with complex tasks, particularly in areas such as mathematics, science and logical reasoning.
A leap in reasoning performance
According to Google, Deep Think significantly improves how the AI tackles difficult problems by running multiple lines of reasoning at the same time. This parallel approach allows it to explore solutions more deeply and accurately than previous versions.
Google claims the new mode delivers standout results on demanding benchmarks. It reportedly achieves 41.0% on Humanity’s Last Exam without the use of tools, and 45.1% on the ARC-AGI-2 test when code execution is enabled. These results, shared via an official Google blog post, are described as setting a new bar for AI-driven reasoning.
