xAI appears to be closing in on two major feature rollouts for Grok on the web—Voice Mode and Tasks, both of which are now functional under feature flags. Voice Mode, which previously wasn’t operational, now supports spoken input and audio responses. It mirrors the mobile version, where users can either type or speak and receive voice replies. What’s especially notable is that some users report hearing replies generated by Grok 3.5, though inconsistently. These anomalies resemble earlier mobile glitches, suggesting that voice capability with Grok 3.5 might still be in limited A/B testing or affected by unstable routing behind the scenes.
BREAKING 🚨: xAI is testing Grok 3.5 along with a voice mode on the Grok web! Sound on.
— TestingCatalog News 🗞 (@testingcatalog) June 12, 2025
Release soon? 👀 pic.twitter.com/luyfmVEh2Z
Tasks, the second feature now being trialed, allows users to schedule either standard Grok queries or “deep search” tasks. The interface supports task creation, editing, pausing, and archiving, alongside a task execution history log, making it easier to track if and when a task ran.

Deep search results are more thorough and are paired with multi-modal notification options: app alerts, emails, or both. Unlike traditional newsletters, email summaries provide a condensed preview, helping users quickly decide whether to open the full report in Grok’s web interface.
xAI cooked with Grok Tasks - users will receive a complete summarised response in the email.
— TestingCatalog News 🗞 (@testingcatalog) June 12, 2025
Best for “Morning AI news” 👀 https://t.co/bYPUjGQc3D pic.twitter.com/9S3kvdLrdX
This system is particularly useful for users seeking recurring research or monitoring workflows, such as daily news scans or tracking specific topics. However, the feature’s new usage limits are far stricter than before:
- 3 recurring daily tasks
- 10 occasional ones
This is down from ten daily and thirty occasional tasks, pointing to either infrastructure tuning or a shift toward tiered access.

Grok 3.5 itself, which many anticipated by now, remains in limbo. The delay might be tied to deeper integration work with these new features. Given that the Tasks interface now offers deep search toggles, the long-awaited Grok 3.5 may play a pivotal role in powering those advanced results. Whether these tools will be locked behind the Supergrok subscription is yet unconfirmed, but for now, both features mark a strong move toward agent-like automation in xAI’s Grok ecosystem.