iOS 26.4: Siri Renewed with LLM and Gemini

New LLM-Based Siri Architecture

With the iOS 26.4 release planned for spring 2026, Apple is making a radical change to the infrastructure of its voice assistant Siri. The current keyword-based detection system is being replaced by a generative AI architecture built on Large Language Models (LLMs). This update aims to fundamentally change how Siri processes commands and its capacity to detect user intent.

Google Gemini and Hybrid Processing

The new infrastructure is powered by a custom model Apple developed in collaboration with the Google Gemini team. This model will offer a more natural interaction by analyzing the context and intent behind what the user says. Maintaining its sensitivity on data privacy, Apple will implement a hybrid solution by distributing processing between on-device and Private Cloud Compute infrastructure.

Technical Capabilities and Automation

With the renewed architecture, Siri will be able to go beyond simple commands and perform more complex automation tasks. The core capabilities expected to arrive with iOS 26.4 are as follows:

Contextual Awareness: Generating responses by analyzing the user's previous activities and personal data.
Screen Awareness: Detecting the content open on the screen and taking action accordingly.
Cross-App Integration: Performing multi-step operations with a single command (for example, editing a photo and sending it as a message to a specific person).

Development Process and Reasons for the Delay

These features, initially planned for release with iOS 18, are reportedly delayed due to failures in the hybrid model that tried to combine the old architecture with the new LLM system. According to statements by Apple's Head of Software Engineering Craig Federighi, the restructuring of the AI team and the reconsideration of the architecture from scratch resulted in a more extensive upgrade than expected. The beta process is expected to begin in the coming months.