NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
npm
npm, #4219 most downloaded on npm
The Firebase AI SDK
Latest stable
3.0.0
Last release 12 days ago
10 Nov 2026
Ships fairly regularly
a new release about every 3 weeks
Nearly every release is documented
notes for 26 of 26 stable releases
Nothing withdrawn
no release was ever pulled
2 years old
255 releases · first in 2025
Drag across the ruler to compare, or tap a release to open it.
Nothing published for this version
`f83cc32` #10392 - Removed APIs that were previously deprecated.
42ce99c #10380 - Refactored TemplateGenerativeModel.generateContent and TemplateGenerativeModel.generateContentStream to accept a unified TemplateRequest object ({ templateId, templateVariables, toolConfig? }) as the first parameter.
Made templateVariables required in both TemplateRequest and StartTemplateChatParams.
f83cc32 #10392 - Removed APIs that were previously deprecated.
Removed the previously deprecated VertexAIBackend (and BackendType.VERTEX_AI). Use AgentPlatformBackend (and BackendType.AGENT_PLATFORM) instead.
The only difference for AgentPlatformBackend is the default
location for accessing the model.
The default location for AgentPlatformBackend is global, whereas the default location for VertexAIBackend was us-central1. To use us-central1 with AgentPlatformBackend, specify getAI(app, { backend: new AgentPlatformBackend('us-central1') }) when initializing the SDK. However, note that most new Gemini models do not support us-central1.
Updated default hybrid-in-cloud model to gemini-3.5-flash-lite.
b70fa6f #10398 - Refactored Part into a discriminated union with explicit type properties, introduced UnknownPart for raw wire and user-provided inputs, stripped type discriminators on wire egress, and ensured defensive deep copying for chat history.
87adc59 #10362 - Deprecated candidateCount, temperature, topP, topK, presencePenalty, and frequencyPenalty in both GenerationConfig and LiveGenerationConfig. These parameters are unsupported in Gemini 3.x and later models.
565da23 #10334 - Exported a ChatSession interface instead of the ChatSession class.
Cleaned up AI Logic doc comments.
Your coding agent can read these notes before it upgrades. Set up the MCP server →
The way you install and initialize the SDK has changed and several APIs have been renamed. When you decide to upgrade, please see the Web Migration Guide for detail.
The way you install and initialize the Node.js has also changed. APIs have been renamed and you now authenticate using a service account instead of a secret or custom token. When you choose to upgrade, please see the Database Server SDK Setup Guide and use the Web Migration Guide to help you update your Node.js code to the latest APIs.
The way you install and initialize the SDK has changed and most APIs have been renamed. When you decide to upgrade, please see the Web Migration Guide for details.
Node.js server SDK:
Token minting and verification for Firebase Authentication is now built into the server SDKs themselves instead of in separate token minting libraries.
All of the existing authentication methods have been removed. Instead, use the databaseAuthVariableOverride option (Node.js) or when initializing the SDK. See Installation & Setup for Servers for more details.
Use the Web Migration Guide or to help you update your Node.js or Java code to the latest APIs.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
`6b17ef5` #10319 - Fixed the issue where the SDK hardcodes invalid 'function' role, causing 400 errors during function calling on gemini-3.6-flash
Nothing published for this version
Nothing published for this version
`d4b83fd` #10267 - Imagen methods and types have been removed due to Imagen models being shut down. As a replacement, you can migrate your apps to use
d4b83fd #10267 - Imagen methods and types have been removed due to Imagen models being shut down. As a replacement, you can migrate your apps to use Gemini Image models (the "Nano Banana" models).Nothing published for this version
`178daf7` #10184 - Deprecated VertexAIBackend in favor of AgentPlatformBackend to reflect the renaming of Vertex AI to Gemini Enterprise Agent Platfor…
7c5e6e5 #10145 - Added support for text-to-speech capabilities (both single- and multi-speaker configurations) for non-Live API models (for example, Gemini TTS models).
178daf7 #10184 - Deprecated VertexAIBackend in favor of AgentPlatformBackend to reflect the renaming of Vertex AI to Gemini Enterprise Agent Platform.
The only difference for AgentPlatformBackend is the default location for accessing the model. The default location for AgentPlatformBackend is global, whereas the default location for VertexAIBackend is us-central1. To use us-central1 with AgentPlatformBackend, specify getAI(app, { backend: new AgentPlatformBackend('us-central1') }) when initializing the SDK. However, note that most new Gemini models do not support us-central1.
736fa3f]:
Nothing published for this version
Nothing published for this version
Nothing published for this version
`9bef5ab` #10035 - Deprecated GoogleMaps.enableWidget, following the announcement by the Grounding for Google Maps service.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
`f0752a0` #9800 - Add initializeDeviceModel() method to allow explicit initialization of on-device model in hybrid mode.
Nothing published for this version
Nothing published for this version
`ffa39f6` #9795 - Added LiveSession.resumeSession() to allow resuming a previous LiveSession. Also added contextWindowCompression feature.
ffa39f6 #9795 - Added LiveSession.resumeSession() to allow resuming a previous LiveSession. Also added contextWindowCompression feature.
86dc0db #9819 - Added support for ImageConfig (aspect ratio and size).
Expanded FinishReason values to include all currently available values provided by the models.
345c5f6 #9458 - AI Logic : Feature : Added support for Grounding with Google Maps.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
`23ab5b9` #9817 - Update code execution and URL context features to reflect that they are generally available and no longer public preview.
Nothing published for this version
`cccb6d0` #9771 - [deprecated] All Imagen models are deprecated and will shut down as early as June 2026. As a replacement, you can migrate your apps…
cccb6d0 #9771 - [deprecated] All Imagen models are deprecated and will shut down as early as June 2026. As a replacement, you can migrate your apps to use Gemini Image models (the "Nano Banana" models).
6cbe865 #9816 - Deprecate topK and temperature properties for hybrid inference mode.
9c8e864 #9763 - Add startChat() for TemplateGenerativeModel.
9c8e864 #9763 - Add automatic function calling and chat history for server prompt templates.
Nothing published for this version
Nothing published for this version
Nothing published for this version
`843a8d7` #9693 (fixes #9625) - Added responseJsonSchema to GenerationConfig.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
`47f8521` #9489 - Added automatic function calling capability when using ChatSession.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
`8123231` #9461 - AI Logic: Extend the LiveSession.receive() return types to include a LiveServiceGoingAwayNotice which signifies that the service wil
8123231 #9461 - AI Logic: Extend the LiveSession.receive() return types to include a LiveServiceGoingAwayNotice
which signifies that the service will soon close the session.
f9254b6 #9475 - Remove "preview" tags from code execution and URL context documentation.
2e74329 #9462 - Add params on UsageMetadata that provide cache-related usage data when the implicit caching feature is used.
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per month.