@ai-sdk/xai 4.0.36 → 4.0.38

package/CHANGELOG.md

@@ -1,5 +1,17 @@
 # @ai-sdk/xai
 
+## 4.0.38
+
+### Patch Changes
+
+- 484293f: feat(xai): support the priority service tier on chat and responses
+
+## 4.0.37
+
+### Patch Changes
+
+- a4d386d: feat(xai): add the Grok 4.6 model IDs and support its `xhigh` reasoning effort
+
 ## 4.0.36
 
 ### Patch Changes

package/dist/index.d.ts

@@ -3,16 +3,21 @@ import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { InferSchema, FetchFunction, WebSocketConstructor } from '@ai-sdk/provider-utils';
 import { ProviderV4, LanguageModelV4, ImageModelV4, Experimental_VideoModelV4, Experimental_RealtimeFactoryV4, SpeechModelV4, TranscriptionModelV4, FilesV4, Experimental_RealtimeModelV4, Experimental_RealtimeModelV4ClientSecretOptions, Experimental_RealtimeModelV4ClientSecretResult, Experimental_RealtimeModelV4ServerEvent, Experimental_RealtimeModelV4ClientEvent, Experimental_RealtimeModelV4SessionConfig } from '@ai-sdk/provider';
 
-type XaiChatModelId = 'grok-4.20-non-reasoning' | 'grok-4.20-reasoning' | 'grok-4.3' | 'grok-4.5' | 'grok-latest' | (string & {});
+type XaiChatModelId = 'grok-4.20-non-reasoning' | 'grok-4.20-reasoning' | 'grok-4.3' | 'grok-4.5' | 'grok-4.6' | 'grok-latest' | (string & {});
 declare const xaiLanguageModelChatOptions: z.ZodObject<{
     reasoningEffort: z.ZodOptional<z.ZodEnum<{
         none: "none";
         low: "low";
         medium: "medium";
         high: "high";
+        xhigh: "xhigh";
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     parallel_function_calling: z.ZodOptional<z.ZodBoolean>;
     searchParameters: z.ZodOptional<z.ZodObject<{
         mode: z.ZodEnum<{
@@ -72,7 +77,7 @@ declare const xaiFilePartProviderOptions: z.ZodObject<{
 }, z.core.$strip>;
 type XaiFilePartProviderOptions = z.infer<typeof xaiFilePartProviderOptions>;
 
-type XaiResponsesModelId = 'grok-4.20-non-reasoning' | 'grok-4.20-reasoning' | 'grok-4.3' | 'grok-4.5' | 'grok-latest' | (string & {});
+type XaiResponsesModelId = 'grok-4.20-non-reasoning' | 'grok-4.20-reasoning' | 'grok-4.3' | 'grok-4.5' | 'grok-4.6' | 'grok-latest' | (string & {});
 /**
  * @see https://docs.x.ai/docs/api-reference#create-new-response
  */
@@ -82,6 +87,7 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
         low: "low";
         medium: "medium";
         high: "high";
+        xhigh: "xhigh";
     }>>;
     reasoningSummary: z.ZodOptional<z.ZodEnum<{
         auto: "auto";
@@ -90,6 +96,10 @@ declare const xaiLanguageModelResponsesOptions: z.ZodObject<{
     }>>;
     logprobs: z.ZodOptional<z.ZodBoolean>;
     topLogprobs: z.ZodOptional<z.ZodNumber>;
+    serviceTier: z.ZodOptional<z.ZodEnum<{
+        default: "default";
+        priority: "priority";
+    }>>;
     store: z.ZodOptional<z.ZodBoolean>;
     previousResponseId: z.ZodOptional<z.ZodString>;
     include: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{

package/dist/index.js

@@ -305,15 +305,17 @@ var xaiLanguageModelChatOptions = z2.object({
    * - `low` (default): Uses some reasoning tokens, but still fast.
    * - `medium`: More thinking for less-latency-sensitive applications.
    * - `high`: Uses more reasoning tokens for deeper thinking.
+   * - `xhigh`: Uses the most reasoning tokens (supported by `grok-4.6`).
    *
    * Note: Not every Grok model accepts every value. Refer to xAI's docs for
    * the values supported by your selected model.
    *
    * @see https://docs.x.ai/docs/guides/reasoning
    */
-  reasoningEffort: z2.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: z2.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
   logprobs: z2.boolean().optional(),
   topLogprobs: z2.number().int().min(0).max(8).optional(),
+  serviceTier: z2.enum(["default", "priority"]).optional(),
   /**
    * Whether to enable parallel function calling during tool use.
    * When true, the model can call multiple functions in parallel.
@@ -557,7 +559,7 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
             low: "low",
             medium: "medium",
             high: "high",
-            xhigh: "high"
+            xhigh: this.modelId === "grok-4.6" ? "xhigh" : "high"
           },
           warnings
         });
@@ -574,6 +576,8 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
       top_p: topP,
       seed,
       reasoning_effort: reasoningEffort,
+      // scheduling priority
+      service_tier: options.serviceTier,
       // parallel function calling
       parallel_function_calling: options.parallel_function_calling,
       // response format
@@ -708,6 +712,11 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
         outputTokens: { total: 0, text: 0, reasoning: 0 }
       },
+      ...response.service_tier != null && {
+        providerMetadata: {
+          xai: { serviceTier: response.service_tier }
+        }
+      },
       request: { body },
       response: {
         ...getResponseMetadata(response),
@@ -776,6 +785,7 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
       raw: void 0
     };
     let usage = void 0;
+    let serviceTier = void 0;
     let isFirstChunk = true;
     const contentBlocks = {};
     const lastReasoningDeltas = {};
@@ -816,6 +826,9 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
             if (value.usage != null) {
               usage = convertXaiChatUsage(value.usage);
             }
+            if (value.service_tier != null) {
+              serviceTier = value.service_tier;
+            }
             const choice = value.choices[0];
             if ((choice == null ? void 0 : choice.finish_reason) != null) {
               finishReason = {
@@ -930,6 +943,9 @@ var XaiChatLanguageModel = class _XaiChatLanguageModel {
                   cacheWrite: 0
                 },
                 outputTokens: { total: 0, text: 0, reasoning: 0 }
+              },
+              ...serviceTier != null && {
+                providerMetadata: { xai: { serviceTier } }
               }
             });
           }
@@ -985,6 +1001,7 @@ var xaiChatResponseSchema = z4.object({
   object: z4.literal("chat.completion").nullish(),
   usage: xaiUsageSchema.nullish(),
   citations: z4.array(z4.string().url()).nullish(),
+  service_tier: z4.string().nullish(),
   code: z4.string().nullish(),
   error: z4.string().nullish()
 });
@@ -1014,7 +1031,8 @@ var xaiChatChunkSchema = z4.object({
     })
   ),
   usage: xaiUsageSchema.nullish(),
-  citations: z4.array(z4.string().url()).nullish()
+  citations: z4.array(z4.string().url()).nullish(),
+  service_tier: z4.string().nullish()
 });
 var xaiStreamErrorSchema = z4.object({
   code: z4.string(),
@@ -1646,7 +1664,8 @@ var xaiResponsesResponseSchema = z7.object({
   object: z7.literal("response"),
   output: z7.array(outputItemSchema),
   usage: xaiResponsesUsageSchema.nullish(),
-  status: z7.string()
+  status: z7.string(),
+  service_tier: z7.string().nullish()
 });
 var xaiResponsesChunkSchema = z7.union([
   z7.object({
@@ -1929,7 +1948,8 @@ var xaiResponsesChunkSchema = z7.union([
     type: z7.literal("response.incomplete"),
     response: z7.object({
       incomplete_details: z7.object({ reason: z7.string() }).nullish(),
-      usage: xaiResponsesUsageSchema.nullish()
+      usage: xaiResponsesUsageSchema.nullish(),
+      service_tier: z7.string().nullish()
     })
   }),
   z7.object({
@@ -1966,14 +1986,16 @@ var xaiLanguageModelResponsesOptions = z8.object({
    * Constrains how hard a reasoning model thinks before responding.
    * Possible values are `none` (disables reasoning entirely; supported by
    * `grok-4.3` and newer reasoning models), `low` (uses fewer reasoning
-   * tokens), `medium`, and `high` (uses more reasoning tokens).
+   * tokens), `medium`, `high` (uses more reasoning tokens), and `xhigh`
+   * (supported by `grok-4.6`).
    *
    * @see https://docs.x.ai/docs/guides/reasoning
    */
-  reasoningEffort: z8.enum(["none", "low", "medium", "high"]).optional(),
+  reasoningEffort: z8.enum(["none", "low", "medium", "high", "xhigh"]).optional(),
   reasoningSummary: z8.enum(["auto", "concise", "detailed"]).optional(),
   logprobs: z8.boolean().optional(),
   topLogprobs: z8.number().int().min(0).max(8).optional(),
+  serviceTier: z8.enum(["default", "priority"]).optional(),
   /**
    * Whether to store the input message(s) and model response for later retrieval.
    * Must be set to `false` for teams with Zero Data Retention (ZDR) enabled,
@@ -2461,7 +2483,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
             low: "low",
             medium: "medium",
             high: "high",
-            xhigh: "high"
+            xhigh: this.modelId === "grok-4.6" ? "xhigh" : "high"
           },
           warnings
         });
@@ -2505,6 +2527,9 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
       },
       ...options.previousResponseId != null && {
         previous_response_id: options.previousResponseId
+      },
+      ...options.serviceTier != null && {
+        service_tier: options.serviceTier
       }
     };
     if (xaiTools2 && xaiTools2.length > 0) {
@@ -2525,7 +2550,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
     };
   }
   async doGenerate(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q;
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r;
     const {
       args: body,
       warnings,
@@ -2713,10 +2738,15 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
         inputTokens: { total: 0, noCache: 0, cacheRead: 0, cacheWrite: 0 },
         outputTokens: { total: 0, text: 0, reasoning: 0 }
       },
-      ...((_q = response.usage) == null ? void 0 : _q.cost_in_usd_ticks) != null && {
+      ...(((_q = response.usage) == null ? void 0 : _q.cost_in_usd_ticks) != null || response.service_tier != null) && {
         providerMetadata: {
           xai: {
-            costInUsdTicks: response.usage.cost_in_usd_ticks
+            ...((_r = response.usage) == null ? void 0 : _r.cost_in_usd_ticks) != null && {
+              costInUsdTicks: response.usage.cost_in_usd_ticks
+            },
+            ...response.service_tier != null && {
+              serviceTier: response.service_tier
+            }
           }
         }
       },
@@ -2763,6 +2793,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
     let hasFunctionCall = false;
     let usage = void 0;
     let costInUsdTicks = void 0;
+    let serviceTier = void 0;
     let isFirstChunk = true;
     const contentBlocks = {};
     const seenToolCalls = /* @__PURE__ */ new Set();
@@ -2776,7 +2807,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
             controller.enqueue({ type: "stream-start", warnings });
           },
           transform(chunk, controller) {
-            var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p;
+            var _a2, _b2, _c2, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q;
             if (options.includeRawChunks) {
               controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
             }
@@ -2907,8 +2938,9 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                 usage = convertXaiResponsesUsage(response2.usage);
                 costInUsdTicks = (_c2 = response2.usage.cost_in_usd_ticks) != null ? _c2 : void 0;
               }
+              serviceTier = (_d = response2.service_tier) != null ? _d : void 0;
               if (event.type === "response.incomplete") {
-                const reason = "incomplete_details" in response2 ? (_d = response2.incomplete_details) == null ? void 0 : _d.reason : void 0;
+                const reason = "incomplete_details" in response2 ? (_e = response2.incomplete_details) == null ? void 0 : _e.reason : void 0;
                 finishReason = {
                   unified: reason ? mapXaiResponsesFinishReason(reason) : "other",
                   raw: reason != null ? reason : "incomplete"
@@ -2922,7 +2954,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
               return;
             }
             if (event.type === "response.failed") {
-              const reason = (_e = event.response.incomplete_details) == null ? void 0 : _e.reason;
+              const reason = (_f = event.response.incomplete_details) == null ? void 0 : _f.reason;
               finishReason = {
                 unified: reason ? mapXaiResponsesFinishReason(reason) : "error",
                 raw: reason != null ? reason : "error"
@@ -3046,13 +3078,13 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                     toolCallId: part.id,
                     toolName,
                     result: {
-                      queries: (_f = part.queries) != null ? _f : [],
-                      results: (_h = (_g = part.results) == null ? void 0 : _g.map((result) => ({
+                      queries: (_g = part.queries) != null ? _g : [],
+                      results: (_i = (_h = part.results) == null ? void 0 : _h.map((result) => ({
                         fileId: result.file_id,
                         filename: result.filename,
                         score: result.score,
                         text: result.text
-                      }))) != null ? _h : null
+                      }))) != null ? _i : null
                     }
                   });
                 }
@@ -3119,17 +3151,17 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                   "x_semantic_search",
                   "x_thread_fetch"
                 ];
-                let toolName = (_i = part.name) != null ? _i : "";
-                if (webSearchSubTools.includes((_j = part.name) != null ? _j : "") || part.type === "web_search_call") {
+                let toolName = (_j = part.name) != null ? _j : "";
+                if (webSearchSubTools.includes((_k = part.name) != null ? _k : "") || part.type === "web_search_call") {
                   toolName = webSearchToolName != null ? webSearchToolName : "web_search";
-                } else if (xSearchSubTools.includes((_k = part.name) != null ? _k : "") || part.type === "x_search_call") {
+                } else if (xSearchSubTools.includes((_l = part.name) != null ? _l : "") || part.type === "x_search_call") {
                   toolName = xSearchToolName != null ? xSearchToolName : "x_search";
                 } else if (part.name === "code_execution" || part.type === "code_interpreter_call" || part.type === "code_execution_call") {
                   toolName = codeExecutionToolName != null ? codeExecutionToolName : "code_execution";
                 } else if (part.type === "mcp_call") {
-                  toolName = (_l = mcpToolName != null ? mcpToolName : part.name) != null ? _l : "mcp";
+                  toolName = (_m = mcpToolName != null ? mcpToolName : part.name) != null ? _m : "mcp";
                 }
-                const toolInput = part.type === "custom_tool_call" ? (_m = part.input) != null ? _m : "" : part.type === "mcp_call" ? (_n = part.arguments) != null ? _n : "" : (_o = part.arguments) != null ? _o : "";
+                const toolInput = part.type === "custom_tool_call" ? (_n = part.input) != null ? _n : "" : part.type === "mcp_call" ? (_o = part.arguments) != null ? _o : "" : (_p = part.arguments) != null ? _p : "";
                 const shouldEmit = part.type === "custom_tool_call" ? event.type === "response.output_item.done" : !seenToolCalls.has(part.id);
                 if (shouldEmit && !seenToolCalls.has(part.id)) {
                   seenToolCalls.add(part.id);
@@ -3190,7 +3222,7 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                           sourceType: "url",
                           id: self.config.generateId(),
                           url: annotation.url,
-                          title: (_p = annotation.title) != null ? _p : annotation.url
+                          title: (_q = annotation.title) != null ? _q : annotation.url
                         });
                       }
                     }
@@ -3245,10 +3277,11 @@ var XaiResponsesLanguageModel = class _XaiResponsesLanguageModel {
                 },
                 outputTokens: { total: 0, text: 0, reasoning: 0 }
               },
-              ...costInUsdTicks != null && {
+              ...(costInUsdTicks != null || serviceTier != null) && {
                 providerMetadata: {
                   xai: {
-                    costInUsdTicks
+                    ...costInUsdTicks != null && { costInUsdTicks },
+                    ...serviceTier != null && { serviceTier }
                   }
                 }
               }
@@ -3687,7 +3720,7 @@ var xaiTools = {
 };
 
 // src/version.ts
-var VERSION = true ? "4.0.36" : "0.0.0-test";
+var VERSION = true ? "4.0.38" : "0.0.0-test";
 
 // src/files/xai-files.ts
 import {