Skip to main content
GemAI supports multiple Google Gemini AI models, each optimized for different use cases. You can select the model that best fits your needs based on speed, capability, and task complexity.

​
Available models

The application provides access to three Gemini models through the AIModel enum:
UserConfig.kt
enum class AIModel(val modelName: String, val number: Int) {
    GEMINI_2_0_FLASH_EXP("gemini-2.0-flash-exp", 0),
    GEMINI_1_5_FLASH("gemini-1.5-flash", 1),
    GEMINI_1_5_PRO("gemini-1.5-pro-latest", 2)
}
Each model has a unique identifier number used for persistence and a model name that corresponds to the Google AI API endpoint.

​
Model characteristics

​
Gemini 2.0 Flash (Experimental)

Model ID: gemini-2.0-flash-expNumber: 0This is the experimental version of Gemini’s latest 2.0 Flash model, offering cutting-edge capabilities and performance improvements.Best for:
  • Testing new AI features
  • Accessing latest model improvements
  • Experimental applications
As an experimental model, behavior may change as Google refines the model. Use this for testing new capabilities before they become stable.

​
Model selection

Models are configured through the UserConfig data class:
UserConfig structure
data class UserConfig(
    val model: AIModel,
    val apiKey: String?,
    val hasApiKey: Boolean
) {
    companion object {
        val DEFAULT = UserConfig(
            model = AIModel.GEMINI_1_5_FLASH,
            apiKey = null,
            hasApiKey = false
        )
    }
}

​
Default configuration

The application defaults to Gemini 1.5 Flash for optimal performance: 
val config = UserConfig.DEFAULT
// model = AIModel.GEMINI_1_5_FLASH

​
Retrieving models from storage

You can convert stored model numbers back to enum values using the fromNum() method:
Model conversion
val modelNumber = 1
val model = AIModel.fromNum(modelNumber)
// Returns: AIModel.GEMINI_1_5_FLASH
The fromNum() method provides safe conversion with fallback:
companion object {
    fun fromNum(id: Int?): AIModel {
        return when (id) {
            0 -> GEMINI_2_0_FLASH_EXP
            1 -> GEMINI_1_5_FLASH
            2 -> GEMINI_1_5_PRO
            else -> GEMINI_1_5_FLASH  // Default fallback
        }
    }
}
If an invalid or null ID is provided, the method defaults to Gemini 1.5 Flash.

​
Model initialization

The application loads the selected model through the BaseAIModel class:
BaseAIModel.kt
abstract class BaseAIModel(private val datastoreRepository: DatastoreRepository) {
    
    protected var modelName: String = AIModel.GEMINI_1_5_FLASH.modelName
        private set
    
    init {
        coroutineScope.launch {
            val deferredModelName = async { datastoreRepository.getModel().modelName }
            modelName = deferredModelName.await()
        }
    }
    
    protected val geminiAIModel: GenerativeModel by lazy {
        getGenerativeModel()
    }
}
Model configuration is loaded asynchronously from the datastore repository during initialization, ensuring the correct model is used for all subsequent requests.

​
Model builder configuration

The actual Google AI model instance is created using the model builder:
Creating the generative model
protected open fun getGenerativeModel(): GenerativeModel {
    return modelBuilder
        .setApiKey(apiKey)
        .setModel(modelName)
        .build()
}

​
Changing models at runtime

To switch models, you need to:
  1. Update user configuration with the new model selection
  2. Save to datastore so it persists across sessions
  3. Reinitialize the AI model instance
val newConfig = userConfig.copy(
    model = AIModel.GEMINI_1_5_PRO
)
datastoreRepository.saveModel(newConfig.model)

​
Model comparison

2.0 Flash Exp

Experimental features, latest capabilities, subject to change

1.5 Flash

Fast responses, balanced quality, recommended for most users

1.5 Pro

Advanced reasoning, highest quality, longer response times

​
API key requirements

All models require a valid Google AI API key to function: 
data class UserConfig(
    val model: AIModel,
    val apiKey: String?,
    val hasApiKey: Boolean
)
The application includes a default API key for testing, but you should configure your own API key in production for better rate limits and control.

​
Model persistence

Your selected model is stored in Android DataStore and automatically loaded on app startup: 
// Retrieve stored model
val storedModel = datastoreRepository.getModel()

// Save new model selection
datastoreRepository.saveModel(AIModel.GEMINI_2_0_FLASH_EXP)
This ensures your model preference persists across app restarts and maintains consistency throughout your chat sessions.

Build docs developers (and LLMs) love

Get started for freeTalk to us