google/models

18 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.

Google language model creation

Language models are created by calling the provider instance with a model ID string, e.g., `google('gemini-2.5-flash')`. The models support tool calls and multi-modal capabilities.

Google forward compatibility with gemini-* model IDs

The provider treats unrecognized `gemini-*` model IDs and `-latest` aliases like the newest supported Gemini generation. Currently this means Gemini 3 request behavior for provider-defined and mixed tools, `thinkingLevel` reasoning, multimodal function responses, and thought signatures. Known legacy Gemini model IDs keep their generation-specific request behavior.

Google model capabilities table

The following Gemini models support Image Input, Object Generation, Tool Usage, Tool Streaming, Google Search, and URL Context: gemini-3.6-flash, gemini-3.5-flash, gemini-3.5-flash-lite, gemini-3.1-pro-preview, gemini-3.1-flash-image-preview, gemini-3.1-flash-lite-preview, gemini-3-pro-preview, gemini-3-pro-image-preview, gemini-3-flash-preview, gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-2.5-flash-lite-preview-06-17, gemini-2.0-flash.

Google realtime models experimental feature

Realtime is an experimental feature. Realtime models are created using `.experimental_realtime()` factory method: `const model = google.experimental_realtime('gemini-3.1-flash-live-preview');`

Google realtime session setup

Realtime sessions run in the browser and require a short-lived token created on the server with `google.experimental_realtime.getToken()`. Example: `const token = await google.experimental_realtime.getToken({ model: 'gemini-3.1-flash-live-preview' });`

Google Live Translation model

The `gemini-3.5-live-translate-preview` model provides low-latency speech-to-speech translation. Configure the target language with `providerOptions.google.translationConfig` containing `targetLanguageCode` (BCP-47 language code, defaults to 'en') and `echoTargetLanguage` (boolean). Input is audio; text input, tools, and custom instructions are not supported.

Google Live Translation example

Example creating Live Translation token: `const token = await google.experimental_realtime.getToken({ model: 'gemini-3.5-live-translate-preview', sessionConfig: { outputModalities: ['audio'], inputAudioTranscription: {}, outputAudioTranscription: {}, providerOptions: { google: { translationConfig: { targetLanguageCode: 'pl', echoTargetLanguage: true } } satisfies GoogleRealtimeModelOptions } } });`

Google translation models experimental feature

Speech translation is an experimental feature. Translation models are created using `.translation()` factory method for server-side streaming: `const model = google.translation('gemini-3.5-live-translate-preview');`

Google translation model with streamTranslate

Translation models are streaming-only and used with `experimental_streamTranslate`. Example: `const result = streamTranslate({ model: google.translation('gemini-3.5-live-translate-preview'), audio: audioStream, inputAudioFormat: { type: 'audio/pcm', rate: 16000 }, targetLanguage: 'es' }); for await (const part of result.fullStream) { if (part.type === 'output-text-delta') { process.stdout.write(part.delta); } }`

Google translation model auto-detection and format

The Gemini Live API for translation auto-detects the source language (sourceLanguage is not supported) and always outputs 24kHz 16-bit PCM audio (outputAudioFormat is not supported). Unsupported settings surface as call warnings.

Google translation model capabilities

The `gemini-3.5-live-translate-preview` model supports: Translated Audio, Translated Text, and Source Transcript.

Google Gemma models available

Gemma models available via Google Generative AI API: gemma-3-27b-it, gemma-3-12b-it

Google embedding models factory method

Google embedding models are created using the `.embedding()` factory method. For example: `google.embedding('gemini-embedding-001')`.

Google embedding model capabilities

Embedding model capabilities: gemini-embedding-001 has 3072 default dimensions, supports custom dimensions, no multimodal support; gemini-embedding-2 has 3072 default dimensions, supports custom dimensions, supports multimodal; gemini-embedding-2-preview has 3072 default dimensions, supports custom dimensions, supports multimodal.

Google image models factory method

Image models are created using the `.image()` factory method. The Google provider supports two types: Imagen models using the `:predict` API, and Gemini image models using the `:generateContent` API.

Google speech models factory method

Speech models are created using the `.speech()` factory method with model id as argument, e.g. `google.speech('gemini-2.5-flash-preview-tts')`.

Google speech model capabilities

Speech model capabilities: gemini-2.5-flash-preview-tts supports multi-speaker and style via instructions; gemini-2.5-pro-preview-tts supports multi-speaker and style via instructions; gemini-3.1-flash-tts-preview supports multi-speaker and style via instructions.

Gemini 3 Pro model ID

The model ID for Gemini 3 Pro is 'gemini-3-pro-preview'.

Give your agent this brain