@ai-sdk/xai 5.0.14 → 5.0.15

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # @ai-sdk/xai
 
+## 5.0.15
+
+### Patch Changes
+
+- 30f32bd: feat(xai): add grok-imagine-image-2.0 image model id
+- 0a84440: feat(xai): add `auto` image quality option
+- 08d7f0a: feat(xai): add `1.5k` image resolution provider option
+- 7c41e41: feat(xai): support transcription model ids, `vadThreshold`, and `opus` audio format
+- Updated dependencies [2959d35]
+- Updated dependencies [59116e6]
+  - @ai-sdk/provider@4.0.22
+  - @ai-sdk/provider-utils@5.0.54
+
 ## 5.0.14
 
 ### Patch Changes

package/dist/index.d.ts

@@ -76,11 +76,13 @@ declare const xaiImageModelOptions: z.ZodObject<{
   sync_mode: z.ZodOptional<z.ZodBoolean>;
   resolution: z.ZodOptional<z.ZodEnum<{
     "1k": "1k";
+    "1.5k": "1.5k";
     "2k": "2k";
   }>>;
   quality: z.ZodOptional<z.ZodEnum<{
     low: "low";
     high: "high";
+    auto: "auto";
     medium: "medium";
   }>>;
   user: z.ZodOptional<z.ZodString>;
@@ -206,8 +208,9 @@ declare const xaiSpeechModelOptionsSchema: import("@ai-sdk/provider-utils").Lazy
 type XaiSpeechModelOptions = InferSchema<typeof xaiSpeechModelOptionsSchema>;
 //#endregion
 //#region src/xai-transcription-model-options.d.ts
+type XaiTranscriptionModelId = 'grok-voice-transcribe-2.0' | (string & {});
 declare const xaiTranscriptionModelOptionsSchema: import("@ai-sdk/provider-utils").LazySchema<{
-  audioFormat?: "pcm" | "mulaw" | "alaw" | null | undefined;
+  audioFormat?: "pcm" | "mulaw" | "alaw" | "opus" | null | undefined;
   sampleRate?: 8000 | 16000 | 22050 | 24000 | 44100 | 48000 | null | undefined;
   language?: string | null | undefined;
   format?: boolean | null | undefined;
@@ -216,6 +219,7 @@ declare const xaiTranscriptionModelOptionsSchema: import("@ai-sdk/provider-utils
   diarize?: boolean | null | undefined;
   keyterm?: string | string[] | null | undefined;
   fillerWords?: boolean | null | undefined;
+  vadThreshold?: number | undefined;
   streaming?: {
     interimResults?: boolean | undefined;
     endpointing?: number | undefined;
@@ -235,7 +239,7 @@ declare const xaiFilesOptionsSchema: import("@ai-sdk/provider-utils").LazySchema
 type XaiFilesOptions = InferSchema<typeof xaiFilesOptionsSchema>;
 //#endregion
 //#region src/xai-image-settings.d.ts
-type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
+type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-2.0' | 'grok-imagine-image-pro' | (string & {});
 //#endregion
 //#region src/tool/code-execution.d.ts
 declare const codeExecutionToolFactory: import("@ai-sdk/provider-utils").ProviderExecutedToolFactory<Record<string, never>, {
@@ -570,12 +574,14 @@ interface XaiProvider extends ProviderV4 {
   speechModel(): SpeechModelV4;
   /**
    * Creates an xAI model for speech-to-text transcription.
+   * When `modelId` is omitted, xAI's default transcription model is used.
    */
-  transcription(): TranscriptionModelV4;
+  transcription(modelId?: XaiTranscriptionModelId): TranscriptionModelV4;
   /**
    * Creates an xAI model for speech-to-text transcription.
+   * When `modelId` is omitted, xAI's default transcription model is used.
    */
-  transcriptionModel(): TranscriptionModelV4;
+  transcriptionModel(modelId?: XaiTranscriptionModelId): TranscriptionModelV4;
   /**
    * Returns the xAI files interface for uploading files.
    */
@@ -645,5 +651,5 @@ declare class XaiRealtimeModel implements Experimental_RealtimeModelV4 {
 //#region src/version.d.ts
 export declare const VERSION: string;
 //#endregion
-export { XaiRealtimeModel as Experimental_XaiRealtimeModel, type XaiRealtimeModelConfig as Experimental_XaiRealtimeModelConfig, type XaiErrorData, type XaiFilePartProviderOptions, type XaiFilesOptions, type XaiImageModelOptions, type XaiImageModelOptions as XaiImageProviderOptions, type XaiLanguageModelResponsesOptions, type XaiLanguageModelResponsesOptions as XaiResponsesProviderOptions, type XaiProvider, type XaiProviderSettings, type XaiSpeechModelOptions, type XaiTranscriptionModelOptions, type XaiVideoModelId, type XaiVideoModelOptions, type XaiVideoModelOptions as XaiVideoProviderOptions };
+export { XaiRealtimeModel as Experimental_XaiRealtimeModel, type XaiRealtimeModelConfig as Experimental_XaiRealtimeModelConfig, type XaiErrorData, type XaiFilePartProviderOptions, type XaiFilesOptions, type XaiImageModelOptions, type XaiImageModelOptions as XaiImageProviderOptions, type XaiLanguageModelResponsesOptions, type XaiLanguageModelResponsesOptions as XaiResponsesProviderOptions, type XaiProvider, type XaiProviderSettings, type XaiSpeechModelOptions, type XaiTranscriptionModelId, type XaiTranscriptionModelOptions, type XaiVideoModelId, type XaiVideoModelOptions, type XaiVideoModelOptions as XaiVideoProviderOptions };
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","names":["BatchV4","RealtimeFactoryV4","RealtimeModelV4ClientSecretOptions","RealtimeModelV4ClientSecretResult","RealtimeModelV4ServerEvent","RealtimeModelV4ClientEvent","RealtimeModelV4SessionConfig","RealtimeModelV4"],"sources":["../src/xai-error.ts","../src/xai-file-part-options.ts","../src/responses/xai-responses-language-model-options.ts","../src/xai-image-model-options.ts","../src/xai-video-settings.ts","../src/xai-video-model-options.ts","../src/xai-speech-model-options.ts","../src/xai-transcription-model-options.ts","../src/files/xai-files-options.ts","../src/xai-image-settings.ts","../src/tool/code-execution.ts","../src/tool/image-generation.ts","../src/tool/mcp-server.ts","../src/tool/view-image.ts","../src/tool/view-x-video.ts","../src/tool/web-search.ts","../src/tool/x-search.ts","../src/tool/index.ts","../src/xai-provider.ts","../src/realtime/xai-realtime-model.ts","../src/version.ts"],"mappings":";;;;cAsBa,oBAAkB,EAAA,mBAAA,EAAA;;;;;;;;;;;;GAI7B,EAAA,KAAA;KAEU,eAAe,EAAE,aAAa;;;cCzB7B,4BAA0B,EAAA;;;;;;GAarC,EAAA,KAAA;KAEU,6BAA6B,EAAE,aAClC;;;KCjBG;;;;cAaC,kCAAgC,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2D3C,EAAA,KAAA;KAEU,mCAAmC,EAAE,aACxC;;;cC3EI,sBAAoB,EAAA;;;;;;;;;;;;;;GAO/B,EAAA,KAAA;KAEU,uBAAuB,EAAE,aAAa;;;KCXtC;;;cCIN,kBAAgB,EAAA;;;;;KAIjB,qBAAqB,EAAE,aAAa;cAEnC,gBAAc,EAAA;;;GAGlB,EAAA,KAAA;cAEI,sBAAoB,EAAA;;;;;;GAWxB,EAAA,KAAA;KAEU,mBAAmB,EAAE,aAAa;KAClC,yBAAyB,EAAE,aAAa;UAE1C;EACR;EACA;EACA,aAAa;;EAEb,iBAAiB;;UAGT;;;;EAIR;;UAGQ,gCACA,uBAAuB;;;;EAI/B;;EAEA;;UAGQ,kCAAkC;;;;EAI1C;;EAEA;;UAGQ,wCACA,uBAAuB;;;;EAI/B;;EAEA;;;;EAIA;;UAGQ,kCACA,uBAAuB;EAC/B;EACA;EACA;;EAEA,YAAY;;UAGJ,kCACA,uBAAuB;;;;;EAK/B;EACA;;UAGQ,yCACA,uBAAuB;;;;;EAK/B;EACA;;;;EAIA;;;;;;;;;;;;;;;;KAiBU,uBACR,4BACA,0BACA,4BACA,kCACA,4BACA;;;cC7HS,8DAA2B;;;;;;YAA3B;;KA2DD,wBAAwB,mBAC3B;;;cC5DI,qEAAkC;;;;;;;;;;;;;;;;;KAuFnC,+BAA+B,mBAClC;;;cCxFI,wDAAqB;;;;;;KAetB,kBAAkB,mBAAmB;;;KCtBrC;;;cCQN,2DAAwB,4BAAA;;;;qBAMjB,gBACX,OAAM,kBAAkB,kEAAiC,qBAAA;;;;;;cCerD,6DAA0B;;;;EAMpB;;;;EAKC;;;;;;;;;EAUA;;qBAQA,kBACX,OAAM,kBAAkB,oEAAmC;;;;EAxBjD;;;;EAKC;;;;cCEP,uDAAoB;EAGhB;EACK;EACH;;EAGG;EACG;EACM;EACL;EACL,UAAA;EACM;;qBAQP,YAAa,MAAM,kBAAkB,8DAAwB;EAlBhE;EACK;EACH;;;;cCrCN,uDAAoB,4BAAA;;;;qBAMb,YACX,OAAM,kBAAkB,8DAA6B,qBAAA;;;;;;cCTjD,wDAAqB,4BAAA;;;;;qBAMd,aACX,OAAM,kBAAkB,+DAA8B,qBAAA;;;;;;;cC8BlD,uDAAoB;EAIlB;IACE;IACA;IACA;;IAGA;IACA;;IAGA;IACA;IACA;;EAEI,UAAA;IAAQ;IAAa;;;EAGd;EACC;EACE;EACO;;qBAQlB,YACX,OAAM,kBAAkB,8DAA6B;EA7B/C;IACE;IACA;IACA;;IAGA;IACA;;IAGA;IACA;IACA;;EAEI,UAAA;IAAQ;IAAa;;;;;cC5B7B,qDAAkB;EAGb;EACA,OAAA;IACL;IACA;IACA;IACA;;;EAIgB;EACC;EACR;EACF;EACkB;EACA;;qBAQlB,UAAW,OAAM,kBAAkB,4DAA2B;EAtBhE;EACA,OAAA;IACL;IACA;IACA;IACA;;;;;qBCxBO;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;MAUsrB;MAA4B;;;MAA0E;;;MAAiF;MAAmC;;;;;;;;;;;;;;MAA1N;MAA4B;;;MAA0E;;;MAAiF;MAAmC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;UCK54B,oBAAoB;GAClC,SAAS,sBAAsB;;;;EAKhC,cAAc,SAAS,sBAAsB;;;;EAK7C,YAAY,SAAS,wBAAwB;;;;EAK7C,sBAAsBA;IACpB,MAAM;IACN,OAAO;;;;;EAMT,MAAM,SAAS,kBAAkB;;;;EAKjC,WAAW,SAAS,kBAAkB;;;;EAKtC,MAAM,SAAS,kBAAkB;;;;EAKjC,WAAW,SAAS,kBAAkB;EAEtC,uBAAuBC;;;;EAKvB,UAAU;;;;EAKV,eAAe;;;;EAKf,iBAAiB;;;;EAKjB,sBAAsB;;;;EAKtB,SAAS;;;;EAKT,cAAc;;;;EAKd,mBAAmB;;UAGJ;;;;EAIf;;;;EAKA;;;;EAKA,UAAU;;;;;EAMV,QAAQ;;;;;EAMR,YAAY;;wBAGE,UAAU,UAAS,sBAA2B;qBA0IjD,KAAG;;;KC1QJ;EACV;EACA;EACA,eAAe;EACf,QAAQ;;cAGG,4BAA4BM;WAC9B;WACA;WACA;mBAEQ;EAEL,YAAA,iBAAiB,QAAQ;EAM/B,qBACJ,SAASL,kDACR,QAAQC;EAwCX,mBAAmB;IAAW;IAAe;;IAC3C;IACA;;EAQF,iBAAiB,eAAeC;EAIhC,qBAAqB,OAAOC;EAI5B,mBACE,QAAQC,4CACP;;;;qBC/FQ"}
+{"version":3,"file":"index.d.ts","names":["BatchV4","RealtimeFactoryV4","RealtimeModelV4ClientSecretOptions","RealtimeModelV4ClientSecretResult","RealtimeModelV4ServerEvent","RealtimeModelV4ClientEvent","RealtimeModelV4SessionConfig","RealtimeModelV4"],"sources":["../src/xai-error.ts","../src/xai-file-part-options.ts","../src/responses/xai-responses-language-model-options.ts","../src/xai-image-model-options.ts","../src/xai-video-settings.ts","../src/xai-video-model-options.ts","../src/xai-speech-model-options.ts","../src/xai-transcription-model-options.ts","../src/files/xai-files-options.ts","../src/xai-image-settings.ts","../src/tool/code-execution.ts","../src/tool/image-generation.ts","../src/tool/mcp-server.ts","../src/tool/view-image.ts","../src/tool/view-x-video.ts","../src/tool/web-search.ts","../src/tool/x-search.ts","../src/tool/index.ts","../src/xai-provider.ts","../src/realtime/xai-realtime-model.ts","../src/version.ts"],"mappings":";;;;cAsBa,oBAAkB,EAAA,mBAAA,EAAA;;;;;;;;;;;;GAI7B,EAAA,KAAA;KAEU,eAAe,EAAE,aAAa;;;cCzB7B,4BAA0B,EAAA;;;;;;GAarC,EAAA,KAAA;KAEU,6BAA6B,EAAE,aAClC;;;KCjBG;;;;cAaC,kCAAgC,EAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2D3C,EAAA,KAAA;KAEU,mCAAmC,EAAE,aACxC;;;cC3EI,sBAAoB,EAAA;;;;;;;;;;;;;;;;GAO/B,EAAA,KAAA;KAEU,uBAAuB,EAAE,aAAa;;;KCXtC;;;cCIN,kBAAgB,EAAA;;;;;KAIjB,qBAAqB,EAAE,aAAa;cAEnC,gBAAc,EAAA;;;GAGlB,EAAA,KAAA;cAEI,sBAAoB,EAAA;;;;;;GAWxB,EAAA,KAAA;KAEU,mBAAmB,EAAE,aAAa;KAClC,yBAAyB,EAAE,aAAa;UAE1C;EACR;EACA;EACA,aAAa;;EAEb,iBAAiB;;UAGT;;;;EAIR;;UAGQ,gCACA,uBAAuB;;;;EAI/B;;EAEA;;UAGQ,kCAAkC;;;;EAI1C;;EAEA;;UAGQ,wCACA,uBAAuB;;;;EAI/B;;EAEA;;;;EAIA;;UAGQ,kCACA,uBAAuB;EAC/B;EACA;EACA;;EAEA,YAAY;;UAGJ,kCACA,uBAAuB;;;;;EAK/B;EACA;;UAGQ,yCACA,uBAAuB;;;;;EAK/B;EACA;;;;EAIA;;;;;;;;;;;;;;;;KAiBU,uBACR,4BACA,0BACA,4BACA,kCACA,4BACA;;;cC7HS,8DAA2B;;;;;;YAA3B;;KA2DD,wBAAwB,mBAC3B;;;KC5DG;cAIC,qEAAkC;;;;;;;;;;;;;;;;;;KAgGnC,+BAA+B,mBAClC;;;cCrGI,wDAAqB;;;;;;KAetB,kBAAkB,mBAAmB;;;KCtBrC;;;cCQN,2DAAwB,4BAAA;;;;qBAMjB,gBACX,OAAM,kBAAkB,kEAAiC,qBAAA;;;;;;cCerD,6DAA0B;;;;EAMpB;;;;EAKC;;;;;;;;;EAUA;;qBAQA,kBACX,OAAM,kBAAkB,oEAAmC;;;;EAxBjD;;;;EAKC;;;;cCEP,uDAAoB;EAGhB;EACK;EACH;;EAGG;EACG;EACM;EACL;EACL,UAAA;EACM;;qBAQP,YAAa,MAAM,kBAAkB,8DAAwB;EAlBhE;EACK;EACH;;;;cCrCN,uDAAoB,4BAAA;;;;qBAMb,YACX,OAAM,kBAAkB,8DAA6B,qBAAA;;;;;;cCTjD,wDAAqB,4BAAA;;;;;qBAMd,aACX,OAAM,kBAAkB,+DAA8B,qBAAA;;;;;;;cC8BlD,uDAAoB;EAIlB;IACE;IACA;IACA;;IAGA;IACA;;IAGA;IACA;IACA;;EAEI,UAAA;IAAQ;IAAa;;;EAGd;EACC;EACE;EACO;;qBAQlB,YACX,OAAM,kBAAkB,8DAA6B;EA7B/C;IACE;IACA;IACA;;IAGA;IACA;;IAGA;IACA;IACA;;EAEI,UAAA;IAAQ;IAAa;;;;;cC5B7B,qDAAkB;EAGb;EACA,OAAA;IACL;IACA;IACA;IACA;;;EAIgB;EACC;EACR;EACF;EACkB;EACA;;qBAQlB,UAAW,OAAM,kBAAkB,4DAA2B;EAtBhE;EACA,OAAA;IACL;IACA;IACA;IACA;;;;;qBCxBO;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;MAUsrB;MAA4B;;;MAA0E;;;MAAiF;MAAmC;;;;;;;;;;;;;;MAA1N;MAA4B;;;MAA0E;;;MAAiF;MAAmC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;UCM54B,oBAAoB;GAClC,SAAS,sBAAsB;;;;EAKhC,cAAc,SAAS,sBAAsB;;;;EAK7C,YAAY,SAAS,wBAAwB;;;;EAK7C,sBAAsBA;IACpB,MAAM;IACN,OAAO;;;;;EAMT,MAAM,SAAS,kBAAkB;;;;EAKjC,WAAW,SAAS,kBAAkB;;;;EAKtC,MAAM,SAAS,kBAAkB;;;;EAKjC,WAAW,SAAS,kBAAkB;EAEtC,uBAAuBC;;;;EAKvB,UAAU;;;;EAKV,eAAe;;;;;EAMf,cAAc,UAAU,0BAA0B;;;;;EAMlD,mBAAmB,UAAU,0BAA0B;;;;EAKvD,SAAS;;;;EAKT,cAAc;;;;EAKd,mBAAmB;;UAGJ;;;;EAIf;;;;EAKA;;;;EAKA,UAAU;;;;;EAMV,QAAQ;;;;;EAMR,YAAY;;wBAGE,UAAU,UAAS,sBAA2B;qBA0IjD,KAAG;;;KC7QJ;EACV;EACA;EACA,eAAe;EACf,QAAQ;;cAGG,4BAA4BM;WAC9B;WACA;WACA;mBAEQ;EAEL,YAAA,iBAAiB,QAAQ;EAM/B,qBACJ,SAASL,kDACR,QAAQC;EAwCX,mBAAmB;IAAW;IAAe;;IAC3C;IACA;;EAQF,iBAAiB,eAAeC;EAIhC,qBAAqB,OAAOC;EAI5B,mBACE,QAAQC,4CACP;;;;qBC/FQ"}

package/dist/index.js

@@ -31,19 +31,29 @@ const xaiImageModelOptions = z.object({
 	aspect_ratio: z.string().optional(),
 	output_format: z.string().optional(),
 	sync_mode: z.boolean().optional(),
-	resolution: z.enum(["1k", "2k"]).optional(),
+	resolution: z.enum([
+		"1k",
+		"1.5k",
+		"2k"
+	]).optional(),
 	quality: z.enum([
 		"low",
 		"medium",
-		"high"
+		"high",
+		"auto"
 	]).optional(),
 	user: z.string().optional()
 });
 //#endregion
 //#region src/xai-image-model.ts
+const fileInputModelIds = /* @__PURE__ */ new Set([
+	"grok-imagine-image",
+	"grok-imagine-image-2.0",
+	"grok-imagine-image-pro"
+]);
 var XaiImageModel = class XaiImageModel {
 	get supportsFileInputs() {
-		return this.modelId === "grok-imagine-image" || this.modelId === "grok-imagine-image-pro" ? true : void 0;
+		return fileInputModelIds.has(this.modelId) ? true : void 0;
 	}
 	get supportsMaskInputs() {
 		return this.supportsFileInputs === true ? false : void 0;
@@ -3077,7 +3087,7 @@ const xaiTools = {
 };
 //#endregion
 //#region src/version.ts
-const VERSION = "5.0.14";
+const VERSION = "5.0.15";
 //#endregion
 //#region src/files/xai-files-options.ts
 const xaiFilesOptionsSchema = lazySchema(() => zodSchema(z.looseObject({
@@ -3930,12 +3940,16 @@ const xaiSpeechTimestampsResponseSchema = z.object({
 //#region src/xai-transcription-model-options.ts
 const xaiTranscriptionModelOptionsSchema = lazySchema(() => zodSchema(z.object({
 	/**
-	* Audio encoding for raw, headerless input audio.
+	* Input audio encoding for raw, headerless audio: `pcm`, `mulaw`, or
+	* `alaw`. Container formats such as Ogg/Opus are auto-detected, so do
+	* not set this for them. `opus` is for streaming only, with one raw Opus
+	* packet per chunk.
 	*/
 	audioFormat: z.enum([
 		"pcm",
 		"mulaw",
-		"alaw"
+		"alaw",
+		"opus"
 	]).nullish(),
 	/**
 	* Sample rate of the input audio in Hz.
@@ -3977,6 +3991,11 @@ const xaiTranscriptionModelOptionsSchema = lazySchema(() => zodSchema(z.object({
 	*/
 	fillerWords: z.boolean().nullish(),
 	/**
+	* Voice activity detection threshold between 0 and 1. Defaults to 0.5
+	* for `transcribe` and 0.08 for streaming. 0 disables the filter.
+	*/
+	vadThreshold: z.number().min(0).max(1).optional(),
+	/**
 	* Options for streaming speech-to-text over WebSocket.
 	*/
 	streaming: z.object({
@@ -4026,21 +4045,11 @@ var XaiTranscriptionModel = class XaiTranscriptionModel {
 			schema: xaiTranscriptionModelOptionsSchema
 		});
 		const formData = new FormData();
-		const transcriptionOptions = {
+		appendXaiSttParams(formData, {
+			...getXaiSttParams(this.modelId, xaiOptions),
 			audio_format: xaiOptions?.audioFormat,
-			sample_rate: xaiOptions?.sampleRate,
-			language: xaiOptions?.language,
-			format: xaiOptions?.format,
-			multichannel: xaiOptions?.multichannel,
-			channels: xaiOptions?.channels,
-			diarize: xaiOptions?.diarize,
-			filler_words: xaiOptions?.fillerWords
-		};
-		for (const [key, value] of Object.entries(transcriptionOptions)) if (value != null) formData.append(key, String(value));
-		if (xaiOptions?.keyterm != null) {
-			const keyterms = Array.isArray(xaiOptions.keyterm) ? xaiOptions.keyterm : [xaiOptions.keyterm];
-			for (const keyterm of keyterms) formData.append("keyterm", keyterm);
-		}
+			format: xaiOptions?.format
+		});
 		const blob = audio instanceof Uint8Array ? new Blob([audio]) : new Blob([convertBase64ToUint8Array(audio)]);
 		const fileExtension = mediaTypeToExtension(mediaType);
 		formData.append("file", new File([blob], "audio", { type: mediaType }), `audio.${fileExtension}`);
@@ -4096,12 +4105,13 @@ var XaiTranscriptionModel = class XaiTranscriptionModel {
 			feature: "providerOptions.xai.format",
 			details: "xAI streaming transcription does not support format."
 		});
-		if (xaiOptions?.audioFormat == null && !isKnownInputAudioFormat(options.inputAudioFormat.type)) warnings.push({
+		if (xaiOptions?.audioFormat == null && !inputAudioFormatEncodings.has(options.inputAudioFormat.type)) warnings.push({
 			type: "other",
-			message: `Unrecognized inputAudioFormat.type "${options.inputAudioFormat.type}"; falling back to raw PCM encoding. Use audio/pcm, audio/pcmu, or audio/pcma, or set providerOptions.xai.audioFormat explicitly.`
+			message: `Unrecognized inputAudioFormat.type "${options.inputAudioFormat.type}"; falling back to raw PCM encoding. Use ${[...inputAudioFormatEncodings.keys()].join(", ")}, or set providerOptions.xai.audioFormat explicitly.`
 		});
 		const url = buildXaiStreamingTranscriptionUrl({
 			baseURL: this.config.baseURL ?? "https://api.x.ai/v1",
+			modelId: this.modelId,
 			inputAudioFormat: options.inputAudioFormat,
 			providerOptions: xaiOptions
 		});
@@ -4262,38 +4272,42 @@ function createXaiStreamingTranscriptionStream({ webSocket, url, headers, warnin
 		}
 	});
 }
-function buildXaiStreamingTranscriptionUrl({ baseURL, inputAudioFormat, providerOptions }) {
+function buildXaiStreamingTranscriptionUrl({ baseURL, modelId, inputAudioFormat, providerOptions }) {
 	const url = toWebSocketUrl(`${baseURL}/stt`);
-	appendSearchParam(url, "sample_rate", providerOptions?.sampleRate ?? inputAudioFormat.rate);
-	appendSearchParam(url, "encoding", providerOptions?.audioFormat ?? encodingFromInputAudioFormat(inputAudioFormat.type));
-	appendSearchParam(url, "language", providerOptions?.language);
-	appendSearchParam(url, "diarize", providerOptions?.diarize);
-	appendSearchParam(url, "filler_words", providerOptions?.fillerWords);
-	appendSearchParam(url, "multichannel", providerOptions?.multichannel);
-	appendSearchParam(url, "channels", providerOptions?.channels);
-	appendSearchParam(url, "interim_results", providerOptions?.streaming?.interimResults);
-	appendSearchParam(url, "endpointing", providerOptions?.streaming?.endpointing);
-	appendSearchParam(url, "smart_turn", providerOptions?.streaming?.smartTurn);
-	appendSearchParam(url, "smart_turn_timeout", providerOptions?.streaming?.smartTurnTimeout);
-	if (providerOptions?.keyterm != null) {
-		const keyterms = Array.isArray(providerOptions.keyterm) ? providerOptions.keyterm : [providerOptions.keyterm];
-		for (const keyterm of keyterms) url.searchParams.append("keyterm", keyterm);
-	}
+	const params = getXaiSttParams(modelId, providerOptions);
+	appendXaiSttParams(url.searchParams, {
+		...params,
+		sample_rate: params.sample_rate ?? inputAudioFormat.rate,
+		encoding: providerOptions?.audioFormat ?? inputAudioFormatEncodings.get(inputAudioFormat.type) ?? "pcm",
+		interim_results: providerOptions?.streaming?.interimResults,
+		endpointing: providerOptions?.streaming?.endpointing,
+		smart_turn: providerOptions?.streaming?.smartTurn,
+		smart_turn_timeout: providerOptions?.streaming?.smartTurnTimeout
+	});
 	return url;
 }
-function appendSearchParam(url, key, value) {
-	if (value != null) url.searchParams.set(key, String(value));
-}
-function isKnownInputAudioFormat(type) {
-	return type === "audio/pcm" || type === "audio/pcmu" || type === "audio/pcma";
+function getXaiSttParams(modelId, options) {
+	return {
+		model: modelId || void 0,
+		sample_rate: options?.sampleRate,
+		language: options?.language,
+		multichannel: options?.multichannel,
+		channels: options?.channels,
+		diarize: options?.diarize,
+		filler_words: options?.fillerWords,
+		vad_threshold: options?.vadThreshold,
+		keyterm: options?.keyterm
+	};
 }
-function encodingFromInputAudioFormat(type) {
-	switch (type) {
-		case "audio/pcmu": return "mulaw";
-		case "audio/pcma": return "alaw";
-		default: return "pcm";
-	}
+function appendXaiSttParams(target, params) {
+	for (const [key, value] of Object.entries(params)) for (const item of Array.isArray(value) ? value : [value]) if (item != null) target.append(key, String(item));
 }
+const inputAudioFormatEncodings = /* @__PURE__ */ new Map([
+	["audio/pcm", "pcm"],
+	["audio/pcmu", "mulaw"],
+	["audio/pcma", "alaw"],
+	["audio/opus", "opus"]
+]);
 function channelId(channelIndex) {
 	return channelIndex == null ? void 0 : `channel-${channelIndex}`;
 }
@@ -4366,8 +4380,8 @@ function createXai(options = {}) {
 			fetch: options.fetch
 		});
 	};
-	const createTranscriptionModel = () => {
-		return new XaiTranscriptionModel("", {
+	const createTranscriptionModel = (modelId = "") => {
+		return new XaiTranscriptionModel(modelId, {
 			provider: "xai.transcription",
 			baseURL,
 			headers: getHeaders,