---
title: "Bring any idea to life with Gemini 3 and Firebase AI Logic"
description: "Bring any idea to life with Gemini 3 and Firebase AI Logic"
published: "2025-11-18"
last_updated: "2025-11-18"
authors:
  - "Rosário Fernandes (Developer Relations Engineer)"
  - "Andrew Heard (Software Engineer)"
categories:
  - "Gemini"
  - "AI Logic"
  - "Android"
  - "Flutter"
  - "Unity"
  - "iOS"
  - "Web"
  - "AI"
  - "App Check"
  - "Remote Config"
canonical_url: "https://firebase.blog/posts/2025/11/gemini-3-firebase-ai-logic"
---

# Bring any idea to life with Gemini 3 and Firebase AI Logic

*Published November 18, 2025 by Rosário Fernandes (Developer Relations Engineer), Andrew Heard (Software Engineer)*

---
**[Gemini 3](https://blog.google/products/gemini/gemini-3), our most
intelligent model that helps you bring any idea to life**, is now available for
mobile and web developers to use through [Firebase AI Logic](https://g.dev/ai/logic-gemini-3-announcement) client SDKs! This
gives you **direct access to Gemini 3 Pro preview** via the Gemini API for
developers on the Blaze plan, so you can securely build AI-powered features and
richer app experiences seamlessly in your client apps without needing to do any
server side setup.

### Try out Gemini 3's capabilities 

Firebase AI Logic supports the vast majority of Gemini 3's capabilities,
including **improved thinking, function calling, thought signatures, and the new
default resolution for input media** to give you higher quality output. 

#### Thought signatures

Gemini 3 responses include a `thought_signature` field within any content part
(for example: `text`, `functionCall`, `inlineData`). These signatures are
encrypted representations of the model's internal thought process and are
essential to maintain thought context across turns. The **Firebase AI Logic
client SDKs automatically handle thought signatures for you**, ensuring the
model has access to the thought context from previous turns without requiring
you to do any manual orchestration.

#### New default resolution for input media

Gemini 3 introduces granular control over multimodal vision processing via the
`media_resolution` parameter. While the new default higher resolution improves
the model's ability to read fine text or identify small details, it can increase
token usage and latency. That's why we'll soon be releasing a **configurable
parameter in the client SDKs so that you can control input media resolution**.

#### Thinking levels

We'll also be adding support for configuring Gemini 3's thinking _levels_ to
give you a more intuitive way to configure how much "thinking" the model can do.
But in the meantime, you can still use thinking _budgets_ to reduce latency and
manage costs today.

### Get started with Gemini 3 in Firebase AI Logic today

Firebase AI Logic can help you harness the incredible reasoning powers of Gemini
3 across platforms (Android, Flutter, Web, iOS, and Unity). You can start
experimenting with Gemini 3 Pro preview by specifying `gemini-3-pro-preview` in
your initialization code:

**`AndroidExample.kt`**

```kotlin
val model = Firebase.ai(
  backend = GenerativeBackend.googleAI()
).generativeModel(
  modelName = "gemini-3-pro-preview",
  generationConfig = generationConfig {
    thinkingConfig = thinkingConfig {
      includeThoughts = true
    }
    temperature = 1.0f
  }
)

val response = model.generateContent(
  "You roll two dice. What’s the probability they add up to 7?"
)

response.thoughtSummary?.let {
  println("Thought Summary: $it")
}
response.text?.let {
  println("Answer: $it")
}
```

### Get insight into Gemini 3's performance 

It's important to monitor the costs, usage, and performance of your AI features
so you can understand what normal usage looks like in your production app, and
have peace of mind that you're staying within the thresholds you want.

Firebase provides an
**[AI monitoring dashboard](https://g.dev/firebase/ai-logic-monitoring-console)
that gives you comprehensive visibility into your AI usage** and lets you slice
data by different dimensions - such as model name. In addition to comparing how
Gemini 3 performs vs. previous models, you can track latency, success/failure
rates, and request/response sizes. You can also inspect traces to see your
requests' attributes, inputs, and outputs, which can improve debugging and help
you optimize your AI implementation to deliver the best user experience. 

### Integrate Firebase services for increased security and scale 

Gemini 3 is Google's most secure model yet, having undergone the most
comprehensive set of safety evaluations of any Google AI model to date. On top
of that, Firebase AI Logic seamlessly integrates with other Firebase services to
help you build even more secure and privacy-conscious AI experiences, and scale
with ease. 

For instance, with [Firebase App Check](https://g.dev/firebase/app-check-gemini-3), you can further
safeguard your API resources by **preventing unauthorized access** to your
backend resources. By using [Firebase Remote Config](https://g.dev/firebase/remote-config-gemini-3), you can
**dynamically update** models, prompts, model parameters, feature flags, and A/B
test new features powered by Gemini 3 without requiring users to download an app
update.

### Grow your app with AI 

Firebase can help you quickly and securely transform and grow your app with AI.
Whether that's automating the process to
[generate user reviews](https://firebase.blog/posts/2025/10/adding-ai-powered-reviews)
that boost your app store ranking, or building
[smarter, privacy-first AI features](https://firebase.blog/posts/2025/10/privacy-first-on-device-ai),
Gemini 3 and Firebase AI Logic can unlock new levels of success. Start using
Google's most intelligent model today directly in your [mobile and web apps](https://g.dev/ai/logic-api-reference)!

For more tips on building and growing apps with AI, make sure to follow us on
[X](https://x.com/firebase),
[LinkedIn](https://www.linkedin.com/showcase/firebase/posts), and
[YouTube](https://www.youtube.com/user/Firebase).