What you need to know
- Google is making its newest artificial intelligence model, Gemini Pro, available for integration with Android apps via the Gemini API.
- Gemini Pro is also available starting today in the Google AI Studio and Google Vertex AI.
- Gemini Pro in Android apps will run off-device in Google's data centers, and can be integrated in apps easily in the preview of Android Studio.
Google made the first version of Gemini Pro, its newest artificial intelligence model, available on Wednesday. The company announced the move in a blog post, which explains that the current Gemini Pro version is intended for enterprises and developers and will get better over the next few months. But now, Google is making it possible for Android app developers to integrate Gemini Pro into their own apps.
Google previously allowed developers to use Gemini Nano, a small AI model, in their apps with on-device processing. The Gemini API now allows app developers to use the Gemini Pro model, which uses off-device processing on cloud servers. The company says that with the Google AI SDK, app developers can add Gemini Pro integration to their existing Android apps without needing to create their own backend infrastructure.
Gemini Pro can also be used to help create code in the Google AI Studio and Google Vertex AI programs. The former is the simpler option, while the latter is a more advanced program with "customization of Gemini with full data control." Both are free for now, with up to 60 requests per minute as the initial quota.
Later, when general availability starts early next year, higher requests per minute will become a paid service. Inputs will be charged at a pay-as-you-go rate of $0.00025 per 1,000 characters or $0.0025 per image, and outputs will cost $0.0005 per 1,000 characters.
Google AI Studio is a "streamlined way for developers to integrate the Gemini Pro model, craft prompts, create API keys, and effortlessly transform ideas into AI apps," according to Google. It works by allowing developers to enter prompts in the AI Studio and automatically generate Kotlin code snippets. These snippets of code can then be integrated with an Android app through the Google AI SDK. There is a get code button in the studio that makes this process simple.
However, it is also possible to interface with the new Gemini API directly instead. Google is adding a new project template called Gemini API Stater in the Android Studio program for working with Gemini Pro integration. It's available in preview for now. The new template already includes the code needed to run the Gemini API and will prompt developers to generate an API key in AI Studio when necessary.
The company claims that Gemini is much cheaper than other AI models and could become the primary way Android apps add processing that is more demanding than smartphones can handle on-device.
Google is already looking ahead to Gemini Ultra, which will be its largest and most powerful AI model when it releases early next year. Gemini will also come to Chrome and Firebase in the future. It's already in Google's Bard chatbot for some queries.
