Google has rolled out a series of updates to Firebase and its Gemini model integrations across Apple’s development ecosystem, Android Studio and its own AI tooling, as the company positions Firebase as a central bridge between AI-powered development tools and Google Cloud infrastructure.
The updates span three major developer events – Apple’s Worldwide Developers Conference, Google I/O 2026 and Google Cloud Next 2026 – and collectively represent a push to embed Gemini models and Firebase services deeper into mobile, web and cross-platform app development workflows.
Apple opened its Foundation Models framework to third-party cloud model providers at WWDC, and Google has made Gemini models available through the Firebase Apple SDK.
Starting with iOS 27, macOS 27, iPadOS 27, visionOS 27 and watchOS 27, model providers can implement a new public LanguageModel protocol to provide a common interface for model inference. Google’s integration means cloud-hosted Gemini models can plug directly into the Foundation Models framework using the same API surface as Apple’s on-device model.
For developers already using Apple’s Foundation Models framework, switching to Gemini models requires a single code change: swapping the model instance. The on-device Apple model and cloud-hosted Gemini models sit behind a shared API surface, allowing developers to move between local and cloud inference depending on use case requirements around cost and latency.
The integration is built on Firebase AI Logic, which lets developers integrate Gemini models directly into iOS, macOS, iPadOS and visionOS apps without needing to build and maintain a separate backend server. Firebase App Check provides abuse protection for the service APIs used to access Gemini models.
The integration is available as a preview release.
Google has also worked with Apple to integrate Gemini into Xcode, giving developers the ability to perform multi-step coding tasks during development without leaving the IDE.
Developers can onboard through the Intelligence settings panel in Xcode. Once configured, Gemini provides an agentic experience for reviewing code, fixing bugs and building features.
Authentication is handled differently depending on the developer’s context. Individual developers can obtain a self-serve Gemini API key from Google AI Studio, which includes both a free tier and a paid tier. Enterprise developers can use the Gemini Enterprise Agent Platform for API keys tied to their organisation’s dedicated corporate quotas and data privacy parameters.
At Google I/O 2026, Firebase received a range of updates focused on making the platform available across more AI-powered development environments.
Firebase is now integrated with Google Antigravity 2.0, a standalone desktop app designed for agent interaction. The new onboarding flow features a one-click Firebase setup that installs all necessary components, including Agent Skills and MCP servers.
Agent Skills for Firebase are now integrated by default in Android Studio’s Agent Mode, with no additional setup required. The skills allow agents to set up Firestore and Firebase Authentication, generate code for Firestore and write security rules.
Google has also expanded Agent Skills for Firebase to cover mobile development across Android, iOS and Flutter, in addition to existing web support. The skills now encompass Crashlytics and Remote Config. Crashlytics skills enable debugging assistance inside IDEs, allowing developers to resolve issues without switching to the Crashlytics dashboard.
Agent Skills can also be used with third-party tools including Claude Code and Codex.
Firebase’s integration with Google AI Studio now supports one-click deployment to Cloud Run without requiring payment for the first two Firebase-enabled apps on Google Cloud’s Starter Tier.
Developers can connect apps to Google Workspace data such as Gmail, Docs and Sheets using natural language. Through a “Sign in with Google” flow powered by Firebase Authentication, apps can securely access Google Workspace data to build custom workflows – for example, triaging an inbox and drafting replies, or transforming documents into slide decks.
Firebase-enabled apps can also be exported from Google AI Studio to Antigravity for multi-agent orchestration. The export carries over source code and relevant Firebase context.
Google AI Studio now supports generating Kotlin-based Android apps, with Firebase integrations including Firestore, Firebase Authentication, Firebase App Check and other Firebase products listed as coming soon. A mobile app for Android and iOS is also in development, which will allow developers to build web apps with Firebase backends from mobile devices.
Firebase AI Logic, which lets developers build generative AI features directly into mobile and web client apps without server-side setup, has received several updates across the event cycle.
On the model support front, Firebase AI Logic now supports all Gemini 3.x models. Google has added Grounding with Google Maps to reduce hallucinations through real-time geospatial context, along with support for Nano Banana’s programmatic image control for tailoring aspect ratio and size of generated images.
For the Gemini Live API, Firebase AI Logic now supports session resumption and context compression to enable conversational app features on unreliable networks.
Security updates include a new “template-only” mode for server prompt templates that enforces Firebase AI Logic to only execute prompts stored on the server, ignoring custom prompt instructions sent from client apps. An “authentication-mode” launching soon will enforce valid Firebase Authentication tokens for Gemini calls. Firebase App Check replay attack protection via one-time tokens is also being introduced.
Hybrid inference is now available for iOS, and Android support has expanded to include Gemma 4. Local web inference in Chrome is set to reach general availability, giving developers the ability to run inference locally using on-device models with fallback to cloud-hosted models.
At Cloud Next 2026, Google evolved Firebase Data Connect into Firebase SQL Connect, connecting mobile and web client apps to Cloud SQL for PostgreSQL.
The updated product adds real-time syncing for transforming relational data into live in-app features, offline cache support for responsiveness on low or no connectivity, and native SQL support that replaces the previous GraphQL requirement. Firebase SQL Connect is available via a no-cost trial.
Firestore Enterprise edition has added pipeline operations and a new query engine. The update introduces native full-text search, geospatial queries, relational-style JOIN using subqueries and a data manipulation language.
The full-text search capability integrates Google’s search technology into Firestore, allowing keyword and phrase searches against the live database without external search synchronisation. Geospatial queries enable location-aware applications to find nearby points of interest. JOIN via subqueries allows developers to aggregate data across multiple collections.
Firestore full-text search and geospatial queries are available in preview, while all pipeline operations have reached general availability.
A developer tutorial published by Google demonstrates how to implement live text search in a React web app using Firestore pipelines. The approach uses the .search() pipeline with a documentMatches expression to search across indexed fields, combined with React Suspense and Tanstack Query for managing loading states. A live search variant adds debouncing and React’s useDeferredValue to prevent unnecessary database queries and loading state flicker.
Google has also published a reference implementation showing how Firebase AI Logic can power a real-time cooking assistant using the Gemini Live API with function calling.
The example app, Friendly Meals, establishes a bidirectional stream where a user can point their phone camera at a cooking area and ask questions while cooking. The Gemini Live model processes the video stream, understands intent using the current recipe as context and streams audio responses back.
The implementation uses client-side function calling to enable hands-free actions. When a user verbally asks the assistant to add an ingredient to their shopping list, the model detects the intent, maps it to a registered function and triggers the action within the app, writing the item to a Firestore-backed grocery list.
Both the model name and system prompt are stored in Firebase Remote Config, allowing updates without requiring users to install a new version of the app.
A new Firebase integration with Google Cloud’s Application Design Center offers a unified model for deploying and managing Firebase resources alongside broader Google Cloud infrastructure.
A standardised Firebase Full-Stack App Essentials Template is available in Google’s Application Template catalogue, including Firestore with security rules, Firebase Authentication and Firebase AI Logic.
Firebase A/B Testing is receiving richer targeting capabilities and tighter integration into Firebase Remote Config’s template setup flow. These updates are rolling out gradually.
Crashlytics, which has historically only supported mobile apps, will soon add web support. The web implementation is built on Google Cloud’s Observability Suite, meaning error and trace data for web apps will be stored alongside server-side data in Cloud Logging and Trace.
Firebase Phone Number Verification has reached general availability, supported by more than 10 carriers across six regions. The feature retrieves the phone number assigned to a device’s SIM over cellular or WiFi without sending SMS messages.
Cloud Functions for Firebase now supports Dart as an experimental feature, aimed at Flutter developers who want to use the same programming language across their frontend and backend.
Last Updated on June 20, 2026 by Nick Ross
We will not spam you. You can easily unsubscribe any time. Read our privacy policy.
Please check your inbox or spam folder to confirm your subscription.
Nick Ross is a veteran tech journalist. He is Editor in Chief of SMBtech, an Australian-focused website for technology and business, and High Performance Laptops.
2023 2024 2025 2026 Adobe Ai Agents AMD Artificial Intelligence (A.I.) Asus Business Cars and Automotive Casinos CEO ChatGPT Cloud Cryptocurrency Cybersecurity eCommerce Enterprise Gambling Gaming GenAi Government HealthTech HP HR (Human Resources) Human Capital Management (HCM) Intel Laptops Large Language Models (LLMs) Lenovo Marketing Martech Melbourne Microsoft Monash University MSI Nvidia Payments Research Retail Security SMB (Small-to-Medium Business) Sustainability Ultraportable
Leave a Reply