The Google Gemini API v2.24.0 introduces new features including 'sample_audio' for Voice in the GAOS SDK and the addition of the Voices API resource. It also exposes several components in public GenAI SDKs and supports new syntax for annotations.
This release introduces new features including credential APIs resource, RetrievalCallStep and RetrievalResultStep to the interactions schema, and support for environments files upload and download across SDKs.
This release introduces 'from_environment' support for environment copying and includes several bug fixes, such as adding '@ts-ignore' before optional MCP imports and removing an unused field from FunctionResultDelta.
The Gemini API has added the 'Gemini 3.8 Flash model' to SDKs and updated the descriptions for Flash models.
This release adds support for the audio/webm MIME type, introduces translation_config SDK support for GEAP, and enhances the Interactions API with Video Understanding capabilities. Additionally, it includes several bug fixes related to Content-Length handling and exposes new elements in the Interactions SDK.
This release introduces new features such as 'AudioTranscriptionConfigMode', support for environment files, and the addition of 'TYPE_JPEG2000' to the VideoContent.MimeType enum. Performance improvements include a significant speed increase in model parsing on the Node tokenizer platform.
This release introduces new features including a `mode` enum for audio transcription configurations, additional parameters for AI search and reinforcement tuning, and an IDLE state for live connection status. It also marks the REQUIRES_ACTION status as deprecated and includes a bug fix for examples.
This release adds support for the feature 'gemini-3.7-flash'.
This release introduces new features including 'interaction_status' for LiveServerContent, makes the deferred service tier publicly available on Vertex, and transforms 'speech_config' into a structured object.
This release introduces the 'Gemini Robotics ER 2 Preview' model and enhances the Interaction resource with a new top-level errors array. Additionally, it improves timeout handling in the JS GenAI SDK and adds a new enum value for FinishReason.
This release introduces new features including the addition of an environment resource, support for GoogleMaps Tool grounding types for places and routing, and per-request retry options in the JS GenAI SDK API Client.
This release introduces a new field `custom_vocabulary` to the BidiGenerateContentSetup message, adds a model selector, includes a queued status for interaction proto, and makes ASR fields public in v1main and Interactions.
This release introduces several new features including the addition of AntigravityAgentConfig and history config, support for preserving custom headers for GAOS, and the new Triggers resource in the GoogleGenAI client.
This release introduces several new features including 'environment_id' in network config, 'response_format' and 'Translation_config' in GenerationConfig, and support for Function declarations during Agent Creation. Additionally, it includes a bug fix for interactions to accept both dict and list[dict] for transform.
The Google Gemini API v2.10.0 introduces new features including 'ComputerUse.disabled_safety_policies', usage fields for steps, and parameters for video generation and response format.
This release introduces several new features including 'audioOffset' for VoiceActivity, 'interimInputTranscription' for LiveServerContent, and enhancements to AudioTranscriptionConfig with 'LanguageAuto', 'LanguageHints', and 'adaptationPhrases'. Additionally, it adds 'ServiceTier' to UsageMetadata and exposes new fields in the Computer Use API.
This release introduces several new features to the Google Gemini API, including the addition of the `ServiceTier` to `UsageMetadata`, new fields in the Computer Use API, and enhancements to the interaction API with new response messages and configuration options.
This release introduces several new features including support for Agent Platform MCP in async content generation, a new transcription language code, a TranslationConfig for live translation, and the addition of ReinforcementTuning support in the GenAI SDK with a ValidateReward API method.
This release introduces several new features including the addition of Skill Registry ListSkills and DeleteSkill to the SDK, support for an additional computer_use field in vertex, and the ability to use 'text/csv' as a supported document mime type for the Interaction API.
The Google Gemini API v2.6.0 introduces the `enable_prompt_injection_detection` feature for Computer Use, adds a `budget_exceeded` status, and includes new fields along with the addition of `gemini-3.5-flash`.
The Google Gemini API v1.48.0 introduces a new feature that supports dedicated TextAnnotationDelta for streaming tool responses and includes a bug fix for service_tier enums.
The Google Gemini API v1.47.0 introduces several new features, including custom metadata for FileSearchResult, additional fields in Veo configs, and enhanced support for various MIME types in content interactions. It also adds model status and service tier information to responses.
The Google Gemini API v1.46.0 introduces several breaking changes to the Interactions API, including refactoring of TextContent annotations and renaming of certain fields. Additionally, new features such as support for Google Maps and updated 'signature' fields have been added, along with a bug fix for quoting a key in LiveClientMessage.
Version 1.45.0 of the Google Gemini API introduces new features including the addition of inference_generation_config to EvaluationConfig for Tuning and support for language code in audio transcription config for the Live API in Vertex AI.
The Google Gemini API v1.44.0 introduces a new model for image previews, enhances support for interaction tool signatures, and updates data types based on the discovery document.
The Google Gemini API v1.43.0 introduces new features such as support for the gemini-3.1-pro-preview model, Image Grounding in the GoogleSearch tool, and enhanced image size support, while also implementing a breaking change by changing the interactions media mime type to an enum.
The Google Gemini API v1.42.0 introduces support for multimodal embedding in Gemini Embedding 2.0 and adds support for MaaS models in the Models.embed_content() method of the Vertex AI API.