Major Changes

• f7d4f01: feat(provider): add support for reasoning-file type for files that are part of reasoning

• 776b617: feat(provider): adding new 'custom' content type

• ef992f8: Remove CommonJS exports from all packages. All packages are now ESM-only ("type": "module"). Consumers using require() must switch to ESM import syntax.

• c29a26f: feat(provider): add support for provider references and uploading files as supported per provider

• 8359612: Start v7 pre-release

• 04e9009: chore: make provider implementations code patterns more consistent, including renaming certain exported symbols

  For all externally exported symbols that were renamed, the old names continue to work via deprecated aliases.

• f62681f: feat(provider/xai): make responses api the default

Patch Changes

• 38fc777: Add AI Gateway hint to provider READMEs

• 85735d8: fix(xai): stop emitting additionalProperties flag

• 74d520f: feat: migrate providers to support new top-level reasoning parameter

• 25f1837: feat(xai): add b64_json response format, usage cost tracking, and quality/user parameters for image models

• a0b0a0c: expose costInUsdTicks in responses provider metadata

• aa5a583: chore(provider/xai): remove obsolete Grok 2 models now that they are shut down in their API

• e5bdc8d: fix (provider/xai): handle mid-stream error chunks

• 4799fa5: chore(provider/xai): update provider to use v4 types

• 813851f: fix (provider/xai): add response.incomplete and response.failed streaming event handling

• 2dc2a52: fix reasoning-start dedupe on multi-summary-part responses to prevent xai 400 on continuation requests

• 9f0e36c: trigger release for all packages after provenance setup

• 58a2ad7: fix: more precise default message for tool execution denial

• d20829e: feat(provider/xai): add moderation error, and costInUsdTicks to video model

• 7fc6bd6: Raise minimum supported Node.js version to 22. Supported versions: 22, 24, and 26.

• 0f11f10: add reasoningSummary to responses API provider options

• 23f9d72: deprecate searchParameters (xai live search) in favor of web_search/x_search agent tools

• 0c4c275: trigger initial canary release

• 1293885: surface full xai error detail in APICallError.message instead of falling back to http status text

• 8d87577: fix(xai): support encrypted reasoning round-trip for ZDR

• 9f20868: fix(provider/xai): correct finish reason for tool calls

• 4f91b5d: chore(provider/xai): update Grok 4.20 model IDs to their non-beta versions

• ce769dd: feat(provider): add experimental Realtime API support for voice conversations

  Adds first-class support for realtime (speech-to-speech) APIs:

  • Experimental_RealtimeModelV4 spec in @ai-sdk/provider with normalized event types and factory
  • OpenAI, Google, and xAI realtime provider implementations
  • openai.experimental_realtime() / google.experimental_realtime() / xai.experimental_realtime() work in both server and browser
  • .getToken() static method on each provider for server-side ephemeral token creation
  • experimental_getRealtimeToolDefinitions helper for provider session tool definitions
  • experimental_useRealtime hook in @ai-sdk/react returning UIMessage[] (aligned with useChat), with onToolCall and addToolOutput for client-driven tool execution
  • inputAudioTranscription session config for showing transcribed user audio messages when supported by the provider

• 12115e9: fix reasoning text extraction from content in responses doGenerate

• 9bd6512: feat(provider): change file part data property to be tagged with a type and remove the image part type

• 258c093: chore: ensure consistent import handling and avoid import duplicates or cycles

• b8396f0: trigger initial beta release

• f51c95e: feat(provider/xai): add video extension and reference-to-video (R2V) support

• 90e2d8a: chore: fix unused vars not being flagged by our lint tooling

• f5181ad: feat(provider/xai): support multiple input images for image editing

• b3976a2: Add workflow serialization support to all provider models.

  @ai-sdk/provider-utils: New serializeModel() helper that extracts only serializable properties from a model instance, filtering out functions and objects containing functions. Third-party provider authors can use this to add workflow support to their own models.

  All providers: headers is now optional in provider config types. This is non-breaking — existing code that passes headers continues to work. Custom provider implementations that construct model configs manually can now omit headers, which is useful when models are deserialized from a workflow step boundary where auth is provided separately.

  All provider model classes now include WORKFLOW_SERIALIZE and WORKFLOW_DESERIALIZE static methods, enabling them to cross workflow step boundaries without serialization errors.

• 1dbecd7: feat(provider/xai): add enableImageSearch to the xAI Web Search tool

  The xAI Responses API supports enable_image_search on Web Search tools. xai.tools.webSearch() now accepts enableImageSearch and sends it through to the API as enable_image_search.

• 80e1702: feat(xai): support 'none' and 'medium' reasoning effort for grok-4.3, and curate the model ID autocomplete list

  grok-4.3 accepts reasoning_effort of "none" | "low" | "medium" | "high", where "none" disables reasoning entirely (no thinking tokens) and "medium" provides more thinking for less-latency-sensitive applications.

  • Adds 'none' to the allowed values for providerOptions.xai.reasoningEffort on both the chat (xai()) and responses (xai.responses()) language models.
  • Adds 'medium' to the chat model's reasoningEffort enum (the responses model already supported it).
  • Top-level reasoning: 'medium' now maps to reasoning_effort: 'medium' for the chat model (previously it was coerced to 'low' because 'medium' was not a valid value).

  In addition, the XaiChatModelId and XaiResponsesModelId autocomplete unions have been trimmed to xAI's current model lineup (docs):

  • grok-4.20-non-reasoning
  • grok-4.20-reasoning
  • grok-4.3
  • grok-latest

  Older entries (grok-3*, grok-4, grok-4-0709, grok-4-latest, grok-4-1-fast-*, grok-4-fast-*, grok-code-fast-1, and grok-4.20-multi-agent-0309) have been removed from the autocomplete list. This is not a runtime change — the model ID type is still open ((string & {})), so passing any model ID that the xAI API accepts continues to work; only IDE autocomplete is affected.

  import { xai } from "@ai-sdk/xai";
  import { generateText } from "ai";
  
  await generateText({
    model: xai("grok-4.3"),
    prompt: "Hi",
    providerOptions: {
      xai: { reasoningEffort: "none" },
    },
  });
  

• 78b6433: feat(provider/xai): support non-image file parts (PDF, text, CSV) in the Responses API via input_file + file_url

  The xAI Responses API accepts { type: 'input_file', file_url } for non-image documents (see https://docs.x.ai/docs/guides/chat-with-files), but the AI SDK xAI Responses provider previously threw UnsupportedFunctionalityError for any file part whose mediaType did not start with image/.

  When a file part is passed with data: URL and a non-image media type, the provider now emits { type: 'input_file', file_url }. application/pdf and text/* are also added to supportedUrls so the SDK does not download them to bytes before reaching the converter.

  Inline-byte (base64) inputs for non-image media types continue to throw, since xAI's Responses API requires either a public URL or a pre-uploaded file_id for non-image documents.

• 7486744: Add xAI speech-to-text transcription support.

• 7486744: feat(provider/xai): add text-to-speech support