Google’s generative model “Gemini AI” in our test
Solutions such as ChatGPT are now firmly established in the field of large language models (LLMs) and generative models. To test the compatibility of such models with the Flutter Framework, we at Bitforge took a closer look at Google’s rival product, “Gemini AI” and its web interface “Google AI Studio”.
Google AI Studio: A user-friendly interface
The Google AI Studio offers an intuitive and user-friendly web interface that makes it easy for users to quickly find their way around. Various services can be used here to obtain the desired output through targeted prompt design.
The Studio makes it possible to refine inputs and further optimize the model using various test cases. There are also numerous setting options available to customize the model. Thanks to the high degree of customization, the Google AI Studio is particularly interesting for developers, as they can design and adapt the model according to their specific ideas.
Intro & demo of the Generative AI features in the Google Cloud
Gemini API: Easy integration into Flutter Apps
The Gemini API provided by Google can be easily integrated into a Flutter app. All you need is a valid API key and the “google_generative_ai” package. As soon as these requirements are met, the API can be used to send queries and process the results in just a few steps.
Create API key: https://ai.google.dev/gemini-api/docs/api-key
Integration google_generative_ai package: https://pub.dev/packages/google_generative_ai/install
Our demo app: Using the Gemini API
To explore the possibilities of the Gemini API, we have developed a demo app. This app uses the Gemini API to show the versatility and customizability of the model within a Flutter app.
We have implemented various use cases in our app, including an interactive chatbot that can take on the personality of different characters and a guessing game in which users have to guess capital cities based on clues. These examples illustrate how effectively the Gemini API can be used in practical applications.
Insight into the demo app
Conclusion: Gemini AI compared to GPT-4
The Gemini model impresses with its simple integration into Flutter apps and the user-friendly interface of Google AI Studio. It is therefore well suited for simple to moderately complex applications, as in the case of our demo app, and offers developers a great deal of flexibility.
Despite its strengths, Gemini AI still has some catching up to do compared to GPT-4, especially for complex scientific and mathematical tasks. GPT-4 remains the preferred choice for more demanding applications.
At Bitforge, we are always on the lookout for innovative solutions to enrich our projects. With Gemini AI, we have found another tool that definitely has potential.
Contribution by: Miguel Seglias, Mobile Developer
