@ai-sdk/xai 4.0.0-beta.0 → 4.0.0-beta.10

package/CHANGELOG.md

@@ -1,5 +1,74 @@
 # @ai-sdk/xai
 
+## 4.0.0-beta.10
+
+### Patch Changes
+
+- Updated dependencies [5c2a5a2]
+  - @ai-sdk/provider@4.0.0-beta.1
+  - @ai-sdk/openai-compatible@3.0.0-beta.4
+  - @ai-sdk/provider-utils@5.0.0-beta.2
+
+## 4.0.0-beta.9
+
+### Patch Changes
+
+- Updated dependencies [8f3e1da]
+  - @ai-sdk/openai-compatible@3.0.0-beta.3
+
+## 4.0.0-beta.8
+
+### Patch Changes
+
+- 4799fa5: chore(provider/xai): update provider to use v4 types
+
+## 4.0.0-beta.7
+
+### Patch Changes
+
+- 0f11f10: add reasoningSummary to responses API provider options
+- 12115e9: fix reasoning text extraction from content in responses doGenerate
+
+## 4.0.0-beta.6
+
+### Major Changes
+
+- f62681f: feat(provider/xai): make responses api the default
+
+## 4.0.0-beta.5
+
+### Patch Changes
+
+- aa5a583: chore(provider/xai): remove obsolete Grok 2 models now that they are shut down in their API
+
+## 4.0.0-beta.4
+
+### Patch Changes
+
+- 25f1837: feat(xai): add b64_json response format, usage cost tracking, and quality/user parameters for image models
+
+## 4.0.0-beta.3
+
+### Patch Changes
+
+- f5181ad: feat(provider/xai): support multiple input images for image editing
+
+## 4.0.0-beta.2
+
+### Patch Changes
+
+- Updated dependencies [45b3d76]
+- Updated dependencies [f7295cb]
+  - @ai-sdk/openai-compatible@3.0.0-beta.2
+
+## 4.0.0-beta.1
+
+### Patch Changes
+
+- Updated dependencies [531251e]
+  - @ai-sdk/provider-utils@5.0.0-beta.1
+  - @ai-sdk/openai-compatible@3.0.0-beta.1
+
 ## 4.0.0-beta.0
 
 ### Major Changes

package/dist/index.d.mts

@@ -1,9 +1,9 @@
 import { z } from 'zod/v4';
-import { ProviderV3, LanguageModelV3, ImageModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { ProviderV4, LanguageModelV4, ImageModelV4, Experimental_VideoModelV4 } from '@ai-sdk/provider';
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 
-type XaiChatModelId = 'grok-4-1-fast-reasoning' | 'grok-4-1-fast-non-reasoning' | 'grok-4-fast-non-reasoning' | 'grok-4-fast-reasoning' | 'grok-code-fast-1' | 'grok-4' | 'grok-4-0709' | 'grok-4-latest' | 'grok-3' | 'grok-3-latest' | 'grok-3-mini' | 'grok-3-mini-latest' | 'grok-2-vision-1212' | 'grok-2-vision' | 'grok-2-vision-latest' | 'grok-2-image-1212' | 'grok-2-image' | 'grok-2-image-latest' | (string & {});
+type XaiChatModelId = 'grok-4-1-fast-reasoning' | 'grok-4-1-fast-non-reasoning' | 'grok-4-fast-non-reasoning' | 'grok-4-fast-reasoning' | 'grok-code-fast-1' | 'grok-4' | 'grok-4-0709' | 'grok-4-latest' | 'grok-3' | 'grok-3-latest' | 'grok-3-mini' | 'grok-3-mini-latest' | (string & {});
 declare const xaiLanguageModelChatOptions: z.ZodObject<{
     reasoningEffort: z.ZodOptional<z.ZodEnum<{
         low: "low";
@@ -68,6 +68,11 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
         high: "high";
         medium: "medium";
     }>>;
+    reasoningSummary: z.ZodOptional<z.ZodEnum<{
+        auto: "auto";
+        concise: "concise";
+        detailed: "detailed";
+    }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
     store: z.ZodOptional<z.ZodBoolean>;
@@ -86,6 +91,12 @@ declare const xaiImageModelOptions: z.ZodObject<{
         "1k": "1k";
         "2k": "2k";
     }>>;
+    quality: z.ZodOptional<z.ZodEnum<{
+        low: "low";
+        high: "high";
+        medium: "medium";
+    }>>;
+    user: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
@@ -99,7 +110,7 @@ type XaiVideoModelOptions = {
     [key: string]: unknown;
 };
 
-type XaiImageModelId = 'grok-2-image' | 'grok-2-image-1212' | 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
+type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
 
 declare const codeExecutionToolFactory: _ai_sdk_provider_utils.ProviderToolFactoryWithOutputSchema<Record<string, never>, {
     output: string;
@@ -300,39 +311,36 @@ declare const xaiTools: {
     }>;
 };
 
-interface XaiProvider extends ProviderV3 {
-    /**
-     * Creates an Xai chat model for text generation.
-     */
-    (modelId: XaiChatModelId): LanguageModelV3;
+interface XaiProvider extends ProviderV4 {
+    (modelId: XaiResponsesModelId): LanguageModelV4;
     /**
      * Creates an Xai language model for text generation.
      */
-    languageModel(modelId: XaiChatModelId): LanguageModelV3;
+    languageModel(modelId: XaiResponsesModelId): LanguageModelV4;
     /**
      * Creates an Xai chat model for text generation.
      */
-    chat: (modelId: XaiChatModelId) => LanguageModelV3;
+    chat: (modelId: XaiChatModelId) => LanguageModelV4;
     /**
-     * Creates an Xai responses model for agentic tool calling.
+     * Creates an Xai responses model for text generation.
      */
-    responses: (modelId: XaiResponsesModelId) => LanguageModelV3;
+    responses: (modelId: XaiResponsesModelId) => LanguageModelV4;
     /**
      * Creates an Xai image model for image generation.
      */
-    image(modelId: XaiImageModelId): ImageModelV3;
+    image(modelId: XaiImageModelId): ImageModelV4;
     /**
      * Creates an Xai image model for image generation.
      */
-    imageModel(modelId: XaiImageModelId): ImageModelV3;
+    imageModel(modelId: XaiImageModelId): ImageModelV4;
     /**
      * Creates an Xai video model for video generation.
      */
-    video(modelId: XaiVideoModelId): Experimental_VideoModelV3;
+    video(modelId: XaiVideoModelId): Experimental_VideoModelV4;
     /**
      * Creates an Xai video model for video generation.
      */
-    videoModel(modelId: XaiVideoModelId): Experimental_VideoModelV3;
+    videoModel(modelId: XaiVideoModelId): Experimental_VideoModelV4;
     /**
      * Server-side agentic tools for use with the responses API.
      */

package/dist/index.d.ts

@@ -1,9 +1,9 @@
 import { z } from 'zod/v4';
-import { ProviderV3, LanguageModelV3, ImageModelV3, Experimental_VideoModelV3 } from '@ai-sdk/provider';
+import { ProviderV4, LanguageModelV4, ImageModelV4, Experimental_VideoModelV4 } from '@ai-sdk/provider';
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { FetchFunction } from '@ai-sdk/provider-utils';
 
-type XaiChatModelId = 'grok-4-1-fast-reasoning' | 'grok-4-1-fast-non-reasoning' | 'grok-4-fast-non-reasoning' | 'grok-4-fast-reasoning' | 'grok-code-fast-1' | 'grok-4' | 'grok-4-0709' | 'grok-4-latest' | 'grok-3' | 'grok-3-latest' | 'grok-3-mini' | 'grok-3-mini-latest' | 'grok-2-vision-1212' | 'grok-2-vision' | 'grok-2-vision-latest' | 'grok-2-image-1212' | 'grok-2-image' | 'grok-2-image-latest' | (string & {});
+type XaiChatModelId = 'grok-4-1-fast-reasoning' | 'grok-4-1-fast-non-reasoning' | 'grok-4-fast-non-reasoning' | 'grok-4-fast-reasoning' | 'grok-code-fast-1' | 'grok-4' | 'grok-4-0709' | 'grok-4-latest' | 'grok-3' | 'grok-3-latest' | 'grok-3-mini' | 'grok-3-mini-latest' | (string & {});
 declare const xaiLanguageModelChatOptions: z.ZodObject<{
     reasoningEffort: z.ZodOptional<z.ZodEnum<{
         low: "low";
@@ -68,6 +68,11 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
         high: "high";
         medium: "medium";
     }>>;
+    reasoningSummary: z.ZodOptional<z.ZodEnum<{
+        auto: "auto";
+        concise: "concise";
+        detailed: "detailed";
+    }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
     store: z.ZodOptional<z.ZodBoolean>;
@@ -86,6 +91,12 @@ declare const xaiImageModelOptions: z.ZodObject<{
         "1k": "1k";
         "2k": "2k";
     }>>;
+    quality: z.ZodOptional<z.ZodEnum<{
+        low: "low";
+        high: "high";
+        medium: "medium";
+    }>>;
+    user: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type XaiImageModelOptions = z.infer<typeof xaiImageModelOptions>;
 
@@ -99,7 +110,7 @@ type XaiVideoModelOptions = {
     [key: string]: unknown;
 };
 
-type XaiImageModelId = 'grok-2-image' | 'grok-2-image-1212' | 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
+type XaiImageModelId = 'grok-imagine-image' | 'grok-imagine-image-pro' | (string & {});
 
 declare const codeExecutionToolFactory: _ai_sdk_provider_utils.ProviderToolFactoryWithOutputSchema<Record<string, never>, {
     output: string;
@@ -300,39 +311,36 @@ declare const xaiTools: {
     }>;
 };
 
-interface XaiProvider extends ProviderV3 {
-    /**
-     * Creates an Xai chat model for text generation.
-     */
-    (modelId: XaiChatModelId): LanguageModelV3;
+interface XaiProvider extends ProviderV4 {
+    (modelId: XaiResponsesModelId): LanguageModelV4;
     /**
      * Creates an Xai language model for text generation.
      */
-    languageModel(modelId: XaiChatModelId): LanguageModelV3;
+    languageModel(modelId: XaiResponsesModelId): LanguageModelV4;
     /**
      * Creates an Xai chat model for text generation.
      */
-    chat: (modelId: XaiChatModelId) => LanguageModelV3;
+    chat: (modelId: XaiChatModelId) => LanguageModelV4;
     /**
-     * Creates an Xai responses model for agentic tool calling.
+     * Creates an Xai responses model for text generation.
      */
-    responses: (modelId: XaiResponsesModelId) => LanguageModelV3;
+    responses: (modelId: XaiResponsesModelId) => LanguageModelV4;
     /**
      * Creates an Xai image model for image generation.
      */
-    image(modelId: XaiImageModelId): ImageModelV3;
+    image(modelId: XaiImageModelId): ImageModelV4;
     /**
      * Creates an Xai image model for image generation.
      */
-    imageModel(modelId: XaiImageModelId): ImageModelV3;
+    imageModel(modelId: XaiImageModelId): ImageModelV4;
     /**
      * Creates an Xai video model for video generation.
      */
-    video(modelId: XaiVideoModelId): Experimental_VideoModelV3;
+    video(modelId: XaiVideoModelId): Experimental_VideoModelV4;
     /**
      * Creates an Xai video model for video generation.
      */
-    videoModel(modelId: XaiVideoModelId): Experimental_VideoModelV3;
+    videoModel(modelId: XaiVideoModelId): Experimental_VideoModelV4;
     /**
      * Server-side agentic tools for use with the responses API.
      */

package/dist/index.js

@@ -18,8 +18,8 @@ var __copyProps = (to, from, except, desc) => {
 var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
 
 // src/index.ts
-var src_exports = {};
-__export(src_exports, {
+var index_exports = {};
+__export(index_exports, {
   VERSION: () => VERSION,
   codeExecution: () => codeExecution,
   createXai: () => createXai,
@@ -31,7 +31,7 @@ __export(src_exports, {
   xai: () => xai,
   xaiTools: () => xaiTools
 });
-module.exports = __toCommonJS(src_exports);
+module.exports = __toCommonJS(index_exports);
 
 // src/xai-provider.ts
 var import_provider7 = require("@ai-sdk/provider");
@@ -370,7 +370,7 @@ function prepareTools({
 // src/xai-chat-language-model.ts
 var XaiChatLanguageModel = class {
   constructor(modelId, config) {
-    this.specificationVersion = "v3";
+    this.specificationVersion = "v4";
     this.supportedUrls = {
       "image/*": [/^https?:\/\/.*$/]
     };
@@ -893,7 +893,9 @@ var xaiImageModelOptions = import_v44.z.object({
   aspect_ratio: import_v44.z.string().optional(),
   output_format: import_v44.z.string().optional(),
   sync_mode: import_v44.z.boolean().optional(),
-  resolution: import_v44.z.enum(["1k", "2k"]).optional()
+  resolution: import_v44.z.enum(["1k", "2k"]).optional(),
+  quality: import_v44.z.enum(["low", "medium", "high"]).optional(),
+  user: import_v44.z.string().optional()
 });
 
 // src/xai-image-model.ts
@@ -901,8 +903,8 @@ var XaiImageModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
     this.config = config;
-    this.specificationVersion = "v3";
-    this.maxImagesPerCall = 1;
+    this.specificationVersion = "v4";
+    this.maxImagesPerCall = 3;
   }
   get provider() {
     return this.config.provider;
@@ -919,7 +921,7 @@ var XaiImageModel = class {
     files,
     mask
   }) {
-    var _a, _b, _c, _d;
+    var _a, _b, _c, _d, _e;
     const warnings = [];
     if (size != null) {
       warnings.push({
@@ -946,22 +948,13 @@ var XaiImageModel = class {
       schema: xaiImageModelOptions
     });
     const hasFiles = files != null && files.length > 0;
-    let imageUrl;
-    if (hasFiles) {
-      imageUrl = (0, import_provider_utils4.convertImageModelFileToDataUri)(files[0]);
-      if (files.length > 1) {
-        warnings.push({
-          type: "other",
-          message: "xAI only supports a single input image. Additional images are ignored."
-        });
-      }
-    }
+    const imageUrls = hasFiles ? files.map((file) => (0, import_provider_utils4.convertImageModelFileToDataUri)(file)) : [];
     const endpoint = hasFiles ? "/images/edits" : "/images/generations";
     const body = {
       model: this.modelId,
       prompt,
       n,
-      response_format: "url"
+      response_format: "b64_json"
     };
     if (aspectRatio != null) {
       body.aspect_ratio = aspectRatio;
@@ -978,8 +971,16 @@ var XaiImageModel = class {
     if ((xaiOptions == null ? void 0 : xaiOptions.resolution) != null) {
       body.resolution = xaiOptions.resolution;
     }
-    if (imageUrl != null) {
-      body.image = { url: imageUrl, type: "image_url" };
+    if ((xaiOptions == null ? void 0 : xaiOptions.quality) != null) {
+      body.quality = xaiOptions.quality;
+    }
+    if ((xaiOptions == null ? void 0 : xaiOptions.user) != null) {
+      body.user = xaiOptions.user;
+    }
+    if (imageUrls.length === 1) {
+      body.image = { url: imageUrls[0], type: "image_url" };
+    } else if (imageUrls.length > 1) {
+      body.images = imageUrls.map((url) => ({ url, type: "image_url" }));
     }
     const baseURL = (_a = this.config.baseURL) != null ? _a : "https://api.x.ai/v1";
     const currentDate = (_d = (_c = (_b = this.config._internal) == null ? void 0 : _b.currentDate) == null ? void 0 : _c.call(_b)) != null ? _d : /* @__PURE__ */ new Date();
@@ -994,11 +995,14 @@ var XaiImageModel = class {
       abortSignal,
       fetch: this.config.fetch
     });
-    const downloadedImages = await Promise.all(
-      response.data.map((image) => this.downloadImage(image.url, abortSignal))
+    const hasAllBase64 = response.data.every((image) => image.b64_json != null);
+    const images = hasAllBase64 ? response.data.map((image) => image.b64_json) : await Promise.all(
+      response.data.map(
+        (image) => this.downloadImage(image.url, abortSignal)
+      )
     );
     return {
-      images: downloadedImages,
+      images,
       warnings,
       response: {
         timestamp: currentDate,
@@ -1009,7 +1013,8 @@ var XaiImageModel = class {
         xai: {
           images: response.data.map((item) => ({
             ...item.revised_prompt ? { revisedPrompt: item.revised_prompt } : {}
-          }))
+          })),
+          ...((_e = response.usage) == null ? void 0 : _e.cost_in_usd_ticks) != null ? { costInUsdTicks: response.usage.cost_in_usd_ticks } : {}
         }
       }
     };
@@ -1028,10 +1033,14 @@ var XaiImageModel = class {
 var xaiImageResponseSchema = import_v45.z.object({
   data: import_v45.z.array(
     import_v45.z.object({
-      url: import_v45.z.string(),
+      url: import_v45.z.string().nullish(),
+      b64_json: import_v45.z.string().nullish(),
       revised_prompt: import_v45.z.string().nullish()
     })
-  )
+  ),
+  usage: import_v45.z.object({
+    cost_in_usd_ticks: import_v45.z.number().nullish()
+  }).nullish()
 });
 
 // src/responses/xai-responses-language-model.ts
@@ -1336,6 +1345,7 @@ var outputItemSchema = import_v46.z.discriminatedUnion("type", [
     type: import_v46.z.literal("reasoning"),
     id: import_v46.z.string(),
     summary: import_v46.z.array(reasoningSummaryPartSchema),
+    content: import_v46.z.array(import_v46.z.object({ type: import_v46.z.string(), text: import_v46.z.string() })).nullish(),
     status: import_v46.z.string(),
     encrypted_content: import_v46.z.string().nullish()
   })
@@ -1642,6 +1652,7 @@ var xaiLanguageModelResponsesOptions = import_v47.z.object({
    * Possible values are `low` (uses fewer reasoning tokens), `medium` and `high` (uses more reasoning tokens).
    */
   reasoningEffort: import_v47.z.enum(["low", "medium", "high"]).optional(),
+  reasoningSummary: import_v47.z.enum(["auto", "concise", "detailed"]).optional(),
   logprobs: import_v47.z.boolean().optional(),
   topLogprobs: import_v47.z.number().int().min(0).max(8).optional(),
   /**
@@ -1951,7 +1962,7 @@ async function prepareResponsesTools({
 // src/responses/xai-responses-language-model.ts
 var XaiResponsesLanguageModel = class {
   constructor(modelId, config) {
-    this.specificationVersion = "v3";
+    this.specificationVersion = "v4";
     this.supportedUrls = {
       "image/*": [/^https?:\/\/.*$/]
     };
@@ -2040,8 +2051,15 @@ var XaiResponsesLanguageModel = class {
           } : { type: "json_object" }
         }
       },
-      ...options.reasoningEffort != null && {
-        reasoning: { effort: options.reasoningEffort }
+      ...(options.reasoningEffort != null || options.reasoningSummary != null) && {
+        reasoning: {
+          ...options.reasoningEffort != null && {
+            effort: options.reasoningEffort
+          },
+          ...options.reasoningSummary != null && {
+            summary: options.reasoningSummary
+          }
+        }
       },
       ...options.store === false && {
         store: options.store
@@ -2070,7 +2088,7 @@ var XaiResponsesLanguageModel = class {
     };
   }
   async doGenerate(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m;
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n;
     const {
       args: body,
       warnings,
@@ -2189,9 +2207,9 @@ var XaiResponsesLanguageModel = class {
           break;
         }
         case "reasoning": {
-          const summaryTexts = part.summary.map((s) => s.text).filter((text) => text && text.length > 0);
-          if (summaryTexts.length > 0) {
-            const reasoningText = summaryTexts.join("");
+          const texts = part.summary.length > 0 ? part.summary.map((s) => s.text) : ((_m = part.content) != null ? _m : []).map((c) => c.text);
+          const reasoningText = texts.filter((text) => text && text.length > 0).join("");
+          if (reasoningText) {
             if (part.encrypted_content || part.id) {
               content.push({
                 type: "reasoning",
@@ -2223,7 +2241,7 @@ var XaiResponsesLanguageModel = class {
       content,
       finishReason: {
         unified: mapXaiResponsesFinishReason(response.status),
-        raw: (_m = response.status) != null ? _m : void 0
+        raw: (_n = response.status) != null ? _n : void 0
       },
       usage: response.usage ? convertXaiResponsesUsage(response.usage) : {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
@@ -2706,7 +2724,7 @@ var xaiTools = {
 };
 
 // src/version.ts
-var VERSION = true ? "4.0.0-beta.0" : "0.0.0-test";
+var VERSION = true ? "4.0.0-beta.10" : "0.0.0-test";
 
 // src/xai-video-model.ts
 var import_provider6 = require("@ai-sdk/provider");
@@ -2737,7 +2755,7 @@ var XaiVideoModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
     this.config = config;
-    this.specificationVersion = "v3";
+    this.specificationVersion = "v4";
     this.maxVideosPerCall = 1;
   }
   get provider() {
@@ -2989,9 +3007,9 @@ function createXai(options = {}) {
       fetch: options.fetch
     });
   };
-  const provider = (modelId) => createChatLanguageModel(modelId);
-  provider.specificationVersion = "v3";
-  provider.languageModel = createChatLanguageModel;
+  const provider = (modelId) => createResponsesLanguageModel(modelId);
+  provider.specificationVersion = "v4";
+  provider.languageModel = createResponsesLanguageModel;
   provider.chat = createChatLanguageModel;
   provider.responses = createResponsesLanguageModel;
   provider.embeddingModel = (modelId) => {