@ai-sdk/google
4.0.122 · 17 Nov 2026Apache-2.014.6M / weekRepositoryHomepagenpmAtom feed
497 stable releases, newest first, by release line. Notes are quoted as the maintainers published them.
4.0.122
17 Nov 2026google: Reject structured output schemas in batch requests because inline Google batch requests ignore
responseJsonSchemaprovider/google: Redact MCP server credentials in telemetry
Google Interactions MCP server headers can contain connector credentials, which
were recorded in tool telemetry.google.tools.mcpServernow provides a
redacted telemetry view of its header values, and Interactions request bodies do
the same for provider error reporting, while the actual request keeps the
credentials.google: Include
promptFeedback.blockReasonMessagein provider metadata
Patch Changes
google: Reject structured output schemas in batch requests because inline Google batch requests ignore
responseJsonSchemaprovider/google: Redact MCP server credentials in telemetry
Google Interactions MCP server headers can contain connector credentials, which
were recorded in tool telemetry.google.tools.mcpServernow provides a
redacted telemetry view of its header values, and Interactions request bodies do
the same for provider error reporting, while the actual request keeps the
credentials.google: Include
promptFeedback.blockReasonMessagein provider metadata
4.0.121
13 Nov 2026Derive known model ID suggestions from each provider's model catalog declarations so catalog updates automatically update autocomplete, while preserving existing suggestions, endpoint-specific model lists, and support for custom model IDs.
Fix Google Interactions ignoring top-level
reasoningin generation and streaming calls. Map reasoning to model-supported thinking levels, including Hydrogen'sxhigh, preserve explicitthinkingLeveloverrides and API defaults, and warn when reasoning is unsupported.
4.0.120
13 Nov 2026Add experimental extraction models backed by structured language model output
Patch Changes
Add experimental extraction models backed by structured language model output
4.0.119
12 Nov 2026Forward the provider's configured fetch transport when downloading file inputs. Add a
fetchoption tocreateDownloadso core downloads can also useguarded-fetchfor opt-in DNS validation on Node.js. The function returned bycreateDownloadcan now also be passed asexperimental_downloadtogenerateTextandstreamText, andexperimental_downloadnow also handles generated file URLs returned by the model, so one hook covers all core downloads. Document that the default URL checks do not resolve DNS and recommend the guarded transport or network egress controls for untrusted URLs.Keep Azure image and Vertex tool-result downloads separate from provider authentication wrappers so file hosts never receive Azure bearer tokens or Vertex Express Mode API keys.
Pass the call's abort signal to custom download requests, including prompt assets and generated files, so
experimental_download: createDownload({ fetch })preserves cancellation and timeouts.Add the
gemini-4-hydrogenmodel ID and support Hydrogen'sxhighthinking level in Google Generate Content and Interactions provider options. Map portablereasoning: 'xhigh'toxhighonly for Hydrogen models.
Patch Changes
Forward the provider's configured fetch transport when downloading file inputs. Add a
fetchoption tocreateDownloadso core downloads can also useguarded-fetchfor opt-in DNS validation on Node.js. The function returned bycreateDownloadcan now also be passed asexperimental_downloadtogenerateTextandstreamText, andexperimental_downloadnow also handles generated file URLs returned by the model, so one hook covers all core downloads. Document that the default URL checks do not resolve DNS and recommend the guarded transport or network egress controls for untrusted URLs.Keep Azure image and Vertex tool-result downloads separate from provider authentication wrappers so file hosts never receive Azure bearer tokens or Vertex Express Mode API keys.
Pass the call's abort signal to custom download requests, including prompt assets and generated files, so
experimental_download: createDownload({ fetch })preserves cancellation and timeouts.Add the
gemini-4-hydrogenmodel ID and support Hydrogen'sxhighthinking level in Google Generate Content and Interactions provider options. Map portablereasoning: 'xhigh'toxhighonly for Hydrogen models.
4.0.118
11 Nov 2026Use the native gemini decisions api for google decision models
Emit File Search and RAG document sources when streaming. Deduplicate by document URI or File Search store and title so distinct documents with matching titles and filenames are preserved.
4.0.117
10 Nov 2026Add support for the
gemini-4-carbonmodel with the same capabilities asgemini-4-argon.
Patch Changes
Add support for the
gemini-4-carbonmodel with the same capabilities asgemini-4-argon.
4.0.116
10 Nov 2026Preserve the served service tier from the
x-gemini-service-tierresponse header when usage metadata omits it, for generated and streamed responses.Apply contiguous array-index validation to dotted numeric paths in streamed tool arguments, while preserving numeric object keys and valid contiguous array updates.
Patch Changes
Preserve the served service tier from the
x-gemini-service-tierresponse header when usage metadata omits it, for generated and streamed responses.Apply contiguous array-index validation to dotted numeric paths in streamed tool arguments, while preserving numeric object keys and valid contiguous array updates.
4.0.114
7 Nov 2026Add
resolveProviderOptionsto parse provider options that a model accepts under severalproviderOptionskeys, either as aliases (first) or merged in order (merge), and use it for the existing multi-key lookups in the OpenAI, OpenAI-compatible, Anthropic, Google, and Amazon Bedrock models. Resolution behavior for these models is unchanged.Pass
providerOptions.google.labelsthrough to Google and Vertex AI Interactions requests, including streaming, so callers can attach request metadata such assafety_identifier.Omit Google provider-defined tools, including File Search and Vertex RAG Store, when
toolChoiceisnoneso disabled tools cannot retrieve data or execute server-side. Preserve function declarations needed by conversation history while disabling function calling. This also applies when structured output uses a synthetic JSON response tool.
Patch Changes
Add
resolveProviderOptionsto parse provider options that a model accepts under severalproviderOptionskeys, either as aliases (first) or merged in order (merge), and use it for the existing multi-key lookups in the OpenAI, OpenAI-compatible, Anthropic, Google, and Amazon Bedrock models. Resolution behavior for these models is unchanged.Pass
providerOptions.google.labelsthrough to Google and Vertex AI Interactions requests, including streaming, so callers can attach request metadata such assafety_identifier.Omit Google provider-defined tools, including File Search and Vertex RAG Store, when
toolChoiceisnoneso disabled tools cannot retrieve data or execute server-side. Preserve function declarations needed by conversation history while disabling function calling. This also applies when structured output uses a synthetic JSON response tool.Skip streamed tool-call array indexes that leave a gap so a small response can't build a huge sparse array
4.0.113
6 Nov 2026Refactor Google Vertex model ID logic into a model catalog that declares capabilities and provider options, and generate provider capability documentation from it. Export Google catalog metadata internally for Vertex to reuse. Request behavior is unchanged.
4.0.112
5 Nov 2026Preserve the order of text and complete function calls in streamed Google responses, including standalone thought signatures on empty text or reasoning parts.
Keep provider metadata on empty text parts when converting results to response messages so it can be passed back in subsequent requests.
Omit foreign-provider empty signature text when replaying history through Anthropic or Bedrock, preserving Bedrock’s existing native reasoning-history handling.
Pass through
seed: 0instead of dropping it in image and video models. KlingAI now reportsseed: 0as an unsupported setting like any other seed.Feat (provider/google): implement server-side realtime extensions
GoogleRealtimeModelsupports the existinggetServerWebSocketConfig()contract, returning the unconstrained Gemini Live endpoint with resolved credential headers. Each server connection can create its own event parser. Reported gross token totals include thinking and tool-use counts, with cached input portions and available modality counts normalized without guessing missing values. Usage in a completion frame is attached toresponse-done; other snapshots usecustomevents withrawType: 'usageMetadata'and typedusage, preserving lifecycle ordering without inferring response associations.
Patch Changes
Preserve the order of text and complete function calls in streamed Google responses, including standalone thought signatures on empty text or reasoning parts.
Keep provider metadata on empty text parts when converting results to response messages so it can be passed back in subsequent requests.
Omit foreign-provider empty signature text when replaying history through Anthropic or Bedrock, preserving Bedrock’s existing native reasoning-history handling.
Pass through
seed: 0instead of dropping it in image and video models. KlingAI now reportsseed: 0as an unsupported setting like any other seed.Feat (provider/google): implement server-side realtime extensions
GoogleRealtimeModelsupports the existinggetServerWebSocketConfig()contract, returning the unconstrained Gemini Live endpoint with resolved credential headers. Each server connection can create its own event parser. Reported gross token totals include thinking and tool-use counts, with cached input portions and available modality counts normalized without guessing missing values. Usage in a completion frame is attached toresponse-done; other snapshots usecustomevents withrawType: 'usageMetadata'and typedusage, preserving lifecycle ordering without inferring response associations.Forward explicit speech language codes to Gemini TTS for single- and multi-speaker requests, including the shared Vertex Gemini adapter. Omitted languages and
language: 'auto'retain automatic detection without an unsupported-language warning.
4.0.111
3 Nov 2026Omit all Google Interactions tools when
toolChoiceisnone, including provider-defined tools.
Patch Changes
Omit all Google Interactions tools when
toolChoiceisnone, including provider-defined tools.
4.0.110
2 Nov 2026google: Handle cycles and BigInt in functionResponse and parse string functionCall arguments
Map Google RPC status names to HTTP status codes when streamed errors omit a numeric code, so transient failures such as
UNAVAILABLEandRESOURCE_EXHAUSTEDare marked retryable.
Patch Changes
google: Handle cycles and BigInt in functionResponse and parse string functionCall arguments
Map Google RPC status names to HTTP status codes when streamed errors omit a numeric code, so transient failures such as
UNAVAILABLEandRESOURCE_EXHAUSTEDare marked retryable.
4.0.109
30 Oct 2026Support promises and synchronous or asynchronous callbacks in provider API key and token settings. Resolve credentials per authenticated request while preserving existing authentication behavior and synchronous credential serialization.
The experimental OpenAI Live
getServerWebSocketConfig()method can now resolve asynchronously. Await its result before using the connection settings.Consolidate Google language model capabilities and request policies into a model catalog shared by runtime resolution and generated documentation, preserving existing Google and Vertex request behavior.
Patch Changes
Support promises and synchronous or asynchronous callbacks in provider API key and token settings. Resolve credentials per authenticated request while preserving existing authentication behavior and synchronous credential serialization.
The experimental OpenAI Live
getServerWebSocketConfig()method can now resolve asynchronously. Await its result before using the connection settings.Consolidate Google language model capabilities and request policies into a model catalog shared by runtime resolution and generated documentation, preserving existing Google and Vertex request behavior.
4.0.108
29 Oct 2026Close string values in tool-call input streamed with
streamFunctionCallArgumentson their final chunkAdd videoMetadata support for file parts. Pass startOffset, endOffset, and fps through per-file providerOptions to control video clipping and frame rate with the Google and Google Vertex providers.
Map the top-level
reasoningoption consistently across providersProviders now resolve reasoning the same way:
mapReasoningToProviderEffortnow accepts anyreasoningvalue, returns an optionalproviderOptionsEffortunchanged, and rounds levels that are missing from the effort map to the closest supported level.mapReasoningToProviderBudgetmaps'none'to the minimum reasoning budget, always returns a budget, and uses the default percentage for levels missing frombudgetPercentages.OpenAICompatibleChatConfigadds an optionalmapReasoningfunction for mapping the top-levelreasoningoption.
Patch Changes
Close string values in tool-call input streamed with
streamFunctionCallArgumentson their final chunkAdd videoMetadata support for file parts. Pass startOffset, endOffset, and fps through per-file providerOptions to control video clipping and frame rate with the Google and Google Vertex providers.
Map the top-level
reasoningoption consistently across providersProviders now resolve reasoning the same way:
A provider-specific reasoning setting in
providerOptionsis sent unchanged. OpenAI no longer dropsproviderOptions.openai.reasoningEffortvalues that are not in its list for the model, and Fireworks no longer rewritesproviderOptions.fireworks.reasoningEffort. Fireworks also sends top-levelreasoningvalues directly instead of rounding them to'high'at most, except for model families known to reject'none'(GPT-OSS, GLM-5.3, Muse Spark) or'minimal'(Nemotron, Inkling).Otherwise,
reasoningmaps to the closest level the model supports, rounding down, and rounds up only when there is no lower level. For example,'xhigh'now maps to'high'instead of'max'on Anthropic and Bedrock models withoutxhigheffort, and'medium'maps to'low'instead of'high'on DeepSeek, Kimi K3, and Mistral GLM-5.3 models.'none'on a model that cannot disable reasoning maps to the model's lowest level with a compatibility warning, instead of being dropped or sending a request the API rejects. This affects GPT-6 models without'none', xAI grok-4.5 and later, Groq GPT-OSS, Kimi K3, Perplexity, QuiverAI, Bedrock OpenAIgpt-ossmodels, and Gemini 2.5 Pro (minimum thinking budget). Gemini 3 and Anthropic and Bedrockbetween_toolsmappings now emit the compatibility warning too.
mapReasoningToProviderEffortnow accepts anyreasoningvalue, returns an optionalproviderOptionsEffortunchanged, and rounds levels that are missing from the effort map to the closest supported level.mapReasoningToProviderBudgetmaps'none'to the minimum reasoning budget, always returns a budget, and uses the default percentage for levels missing frombudgetPercentages.OpenAICompatibleChatConfigadds an optionalmapReasoningfunction for mapping the top-levelreasoningoption.
4.0.106 – 4.0.1072 releases that only update dependencies@ai-sdk/provider-utils 5.0.70Show each
- 4.0.10727 Oct 2026@ai-sdk/provider-utils 5.0.70
- 4.0.10626 Oct 2026@ai-sdk/provider-utils 5.0.69
4.0.105
24 Oct 2026google: Keep empty arrays and objects in tool-call input streamed with
streamFunctionCallArgumentsVertex sends a value-less
partialArgsentry for an empty container (e.g.$.tags), which was dropped, so `` and{}arguments went missing from the streamed tool-call input.
Patch Changes
google: Keep empty arrays and objects in tool-call input streamed with
streamFunctionCallArgumentsVertex sends a value-less
partialArgsentry for an empty container (e.g.$.tags), which was dropped, so `` and{}arguments went missing from the streamed tool-call input.
4.0.104
22 Oct 2026Only updates dependencies:@ai-sdk/provider-utils 5.0.67@ai-sdk/provider 4.0.33
4.0.103
22 Oct 2026Report
interruptedwhen a stream ends cleanly without its provider completion signal, preserving partial output without emitting an error solely for the interruption.This applies to OpenAI (Chat, Responses, and Completions), Anthropic, Google (Generate Content and Interactions), Alibaba, Amazon Bedrock Converse, Cohere, DeepSeek, Groq, Hugging Face Responses, Mistral Chat, Moonshot AI, Open Responses, Perplexity, and xAI Responses. It also covers cutoffs in later Anthropic messages and Google automatic continuations.
Compatibility notice: These streams previously fell back to
otheror omitted the finish part. Applications must now handleinterruptedin finish-reason switches, mappings, and custom validation. CheckfinishReason === 'interrupted'to detect incomplete responses; interruption alone does not trigger error handling.Explicit unknown completion reasons still map to
other. Actual provider, parsing, and transport errors retain their error paths; affected error paths that previously left the finish reason asothernow reporterror. OpenAI-compatible remains unchanged.
Patch Changes
Report
interruptedwhen a stream ends cleanly without its provider completion signal, preserving partial output without emitting an error solely for the interruption.This applies to OpenAI (Chat, Responses, and Completions), Anthropic, Google (Generate Content and Interactions), Alibaba, Amazon Bedrock Converse, Cohere, DeepSeek, Groq, Hugging Face Responses, Mistral Chat, Moonshot AI, Open Responses, Perplexity, and xAI Responses. It also covers cutoffs in later Anthropic messages and Google automatic continuations.
Compatibility notice: These streams previously fell back to
otheror omitted the finish part. Applications must now handleinterruptedin finish-reason switches, mappings, and custom validation. CheckfinishReason === 'interrupted'to detect incomplete responses; interruption alone does not trigger error handling.Explicit unknown completion reasons still map to
other. Actual provider, parsing, and transport errors retain their error paths; affected error paths that previously left the finish reason asothernow reporterror. OpenAI-compatible remains unchanged.
4.0.102
21 Oct 2026Add a shared encodePathSegment helper and use it to encode provider-returned identifiers in credentialed follow-up request paths.
Defer construction of provider validation schemas until first use to reduce package import time.
Patch Changes
Add a shared encodePathSegment helper and use it to encode provider-returned identifiers in credentialed follow-up request paths.
Defer construction of provider validation schemas until first use to reduce package import time.
4.0.101
20 Oct 2026Only updates dependencies:@ai-sdk/provider 4.0.31@ai-sdk/provider-utils 5.0.64
4.0.100
19 Oct 2026Preserve grounding metadata from each automatically continued Google response in
groundingMetadataPerRequest. Previously, a later response without grounding could leave earlier Search queries absent from the final provider metadata even though their sources were returned. The new array preserves request order and includesnullfor requests without grounding.
Patch Changes
Preserve grounding metadata from each automatically continued Google response in
groundingMetadataPerRequest. Previously, a later response without grounding could leave earlier Search queries absent from the final provider metadata even though their sources were returned. The new array preserves request order and includesnullfor requests without grounding.
4.0.99
16 Oct 2026google: Surface in-stream errors
Patch Changes
google: Surface in-stream errors
4.0.98
15 Oct 2026provider/google: Allow AUDIO response modalities for Google language models
4.0.96 – 4.0.972 releases that only update dependencies@ai-sdk/provider 4.0.28@ai-sdk/provider-utils 5.0.61Show each
4.0.95
13 Oct 2026google: Add
gemini-4-argonmodel id and long decode support forgenerateContent: requests that reach the per-request cap (CONTINUATION) are continued automatically with usage summed across requests and the first request's prompt token count exposed asproviderMetadata.google.initialPromptTokenCount, and thecontinuationTokenprovider option / provider metadata resume responses that used up theirmaxOutputTokensbudget.generateTextwith the Interactions API now pollsbackground: truemodel calls until they finish.
4.0.94
12 Oct 2026google: Handle regional Gemini model IDs and valid thinking configurations
Patch Changes
google: Handle regional Gemini model IDs and valid thinking configurations
4.0.93
8 Oct 2026Map the top level embedding dimensions to provider settings
Patch Changes
Map the top level embedding dimensions to provider settings
4.0.92
8 Oct 2026Only updates dependencies:@ai-sdk/provider 4.0.25@ai-sdk/provider-utils 5.0.57
4.0.91
7 Oct 2026Preserve tool errors in OpenAI Chat Completions and Responses by wrapping their payloads in an
errorobject. Send Google tool errors and denied executions underfunctionResponse.response.error.
4.0.90
7 Oct 2026Only updates dependencies:@ai-sdk/provider 4.0.24@ai-sdk/provider-utils 5.0.56
4.0.89
6 Oct 2026Limit jsonl rows to 64mb during batch result downloads
google: Preserve file search custom metadata in retrieved context grounding chunks
google: Add the 'gemini-nano-banana-2.1' model ID
Patch Changes
Limit jsonl rows to 64mb during batch result downloads
google: Preserve file search custom metadata in retrieved context grounding chunks
google: Add the 'gemini-nano-banana-2.1' model ID
provider: Add a portable
maxreasoning level with native provider mappings, compatibility coercions, and budget-based fallback support.Compatibility notes:
TypeScript source compatibility: Adding
maxwidensLanguageModelV4CallOptions['reasoning']and the derivedReasoningLeveltype. Third-party providers and consumers with exhaustive switches orRecord<ReasoningLevel, ...>mappings must handlemaxbefore upgrading. Ordinary calls using existing reasoning levels remain accepted.Amazon Bedrock: Existing Nova 2 calls using portable
reasoning: 'xhigh'now sendmaxReasoningEffort: 'high'instead of the unsupported'max', and return a compatibility warning. Anthropic and OpenAI Bedrock effort mappings retain their existingxhighbehavior. Partial Bedrock reasoning configurations continue to preserve explicit values and derive missing fields; explicit effort/budget overrides and disabled thinking no longer emit warnings for portable mappings that are not sent.Fireworks: Existing portable
minimalandxhighcalls keep their request mappings tolowandhigh, respectively, and now return compatibility warnings in generation and streaming results. Portablemaxmaps tohighwith the same warning behavior.
Release classification: This remains a patch changeset under the repository's explicit release policy, which uses patch releases for both fixes and features. The TypeScript source-compatibility caveat for exhaustive consumers is disclosed above. Maintainers can override this classification with a
majorlabel if they decide to align the provider-spec addition with a future AI SDK major release.
4.0.88
5 Oct 2026Rename
evaluatetodecidegroq,google: Use JSON response tool for structured outputs with tool calling
Patch Changes
Rename
evaluatetodecidegroq,google: Use JSON response tool for structured outputs with tool calling
4.0.87
30 Sep 2026ai: Add telemetry to speech generation and transcription, including provider usage propagation and experimental streaming lifecycle callbacks
Patch Changes
ai: Add telemetry to speech generation and transcription, including provider usage propagation and experimental streaming lifecycle callbacks
4.0.86
30 Sep 2026Only updates dependencies:@ai-sdk/provider 4.0.20@ai-sdk/provider-utils 5.0.52
4.0.85
28 Sep 2026Use standards-compliant User-Agent header
Patch Changes
Use standards-compliant User-Agent header
4.0.84
28 Sep 2026google: Preserve realtime user utterances with independent IDs and cumulative transcripts, honor text-free transcription completion markers, and advance response IDs after interruption
4.0.83
28 Sep 2026google: Send the default Gemini Live
thinkingLevelwhenthinkingConfigsets neitherthinkingLevelnorthinkingBudget, and stop overwriting a rawgenerationConfig.thinkingConfigprovider: Advertise image model file and mask input support
Use confirmed model IDs for capability declarations so unrecognized model names remain unknown. Include Together AI FLUX.2 Pro and Flex single-image editing, and allow asynchronous capability lookups and middleware overrides to resolve to unknown.
Advertise QuiverAI Arrow 2 and Arrow 2 Telos file-input support, and mark Together AI Gemini image inputs unsupported by the current single-image request mapping.
Patch Changes
google: Send the default Gemini Live
thinkingLevelwhenthinkingConfigsets neitherthinkingLevelnorthinkingBudget, and stop overwriting a rawgenerationConfig.thinkingConfigprovider: Advertise image model file and mask input support
Use confirmed model IDs for capability declarations so unrecognized model names remain unknown. Include Together AI FLUX.2 Pro and Flex single-image editing, and allow asynchronous capability lookups and middleware overrides to resolve to unknown.
Advertise QuiverAI Arrow 2 and Arrow 2 Telos file-input support, and mark Together AI Gemini image inputs unsupported by the current single-image request mapping.
4.0.82
25 Sep 2026google: Forward supported Vertex tool result URLs as function response file data
Patch Changes
google: Forward supported Vertex tool result URLs as function response file data
4.0.80
24 Sep 2026google: Serialize JSON Schema references in function responses
4.0.79
23 Sep 2026Fix Google
embedManycalls with more than 100 values by keeping per-value multimodal content aligned across automatic batches, including text-only entries. Validate content length before sending requests and validate each batch's provider options after middleware transforms them.google: Preserve code execution parts when replaying messages
Add Gemini 3.8 TTS support with structured speech metadata and per-turn speaker and style controls for prebuilt voices. Preserve native WAV responses without adding a second header, support explicit raw PCM, mu-law, and A-law output, and identify headerless audio formats correctly. Add the Gemini 3.8 speech model IDs to Google and Gateway types.
Share transcript and custom-voice inspection through the Google provider internal export, and reject empty speech transcripts before sending a request. Default newer and custom model IDs to structured speech while preserving the legacy format for Gemini 2.5 and 3.1.
Patch Changes
Fix Google
embedManycalls with more than 100 values by keeping per-value multimodal content aligned across automatic batches, including text-only entries. Validate content length before sending requests and validate each batch's provider options after middleware transforms them.google: Preserve code execution parts when replaying messages
Add Gemini 3.8 TTS support with structured speech metadata and per-turn speaker and style controls for prebuilt voices. Preserve native WAV responses without adding a second header, support explicit raw PCM, mu-law, and A-law output, and identify headerless audio formats correctly. Add the Gemini 3.8 speech model IDs to Google and Gateway types.
Share transcript and custom-voice inspection through the Google provider internal export, and reject empty speech transcripts before sending a request. Default newer and custom model IDs to structured speech while preserving the legacy format for Gemini 2.5 and 3.1.
4.0.78
23 Sep 2026provider: Preserve opaque file URI strings for provider serialization
Patch Changes
provider: Preserve opaque file URI strings for provider serialization
4.0.77
22 Sep 2026google: Preserve image candidate finish reasons in provider metadata
4.0.76
18 Sep 2026Only updates a dependency:@ai-sdk/provider-utils 5.0.45
4.0.75
18 Sep 2026google: Advertise the supported Gemini image per-call limit
google: Ignore default prompt block reasons in non-streaming responses
google: Include tool-use prompt tokens in input usage
Patch Changes
google: Advertise the supported Gemini image per-call limit
google: Ignore default prompt block reasons in non-streaming responses
google: Include tool-use prompt tokens in input usage
4.0.74
16 Sep 2026Add
google.evaluationModel()for experimental Choice, Score, and Boolean evaluations through Gemini structured output, preserving provider thinking options and validating exact labels and score bounds. Boolean answers contain prompted P(true) estimates validated to be in [0, 1]. Boolean estimates are not guaranteed to be calibrated; application code chooses thresholds.
Patch Changes
Add
google.evaluationModel()for experimental Choice, Score, and Boolean evaluations through Gemini structured output, preserving provider thinking options and validating exact labels and score bounds. Boolean answers contain prompted P(true) estimates validated to be in [0, 1]. Boolean estimates are not guaranteed to be calibrated; application code chooses thresholds.
4.0.73
16 Sep 2026google: Preserve prompt feedback and metadata across streaming chunks
Keep default Node.js downloads protected by DNS validation and connection pinning when frameworks or instrumentation wrap global fetch before or after the SDK loads.
Patch Changes
google: Preserve prompt feedback and metadata across streaming chunks
Keep default Node.js downloads protected by DNS validation and connection pinning when frameworks or instrumentation wrap global fetch before or after the SDK loads.
4.0.72
15 Sep 2026google: Realtime session options for Gemini 3.8 Live
Add
thinkingConfig(thinkingLevel,thinkingBudget,includeThoughts) anddefaultToolBehaviortoGoogleRealtimeModelOptions.thinkingConfigis merged into the Livesetup.generationConfig. Background-reasoning Live models such asgemini-3.8-live-extended-thinkingrequire exactly one ofthinkingLevelorthinkingBudget, so the provider sendsthinkingLevel: 'low'on those models when neither is set.defaultToolBehaviorstampsbehavioron every function declaration in the setup.Forward the Live
interactionStatusandwaitingForInputserver messages as custom events so applications can tell when a background-reasoning model is idle, sinceturnCompletealone no longer means that.
4.0.71
15 Sep 2026Only updates dependencies:@ai-sdk/provider 4.0.15@ai-sdk/provider-utils 5.0.41
A whole project to check? Paste its package.json in the dependency check: every deprecated, outdated or vulnerable dependency, with what to upgrade to. Free, no account.