[evals] adding grok 4 with ai sdk (#890)

# why

benchmarking sota models on stagehand

# what changed

added xai to evals list

# test plan

Author: Kylejeong2

evals/index.eval.ts

@@ -39,6 +39,7 @@ import { groq } from "@ai-sdk/groq";
 import { cerebras } from "@ai-sdk/cerebras";
 import { openai } from "@ai-sdk/openai";
 import { AISdkClient } from "@/examples/external_clients/aisdk";
+import { xai } from "@ai-sdk/xai";
 dotenv.config();
 
 /**
@@ -350,6 +351,10 @@ const generateFilteredTestcases = (): Testcase[] => {
                 ),
               ),
             });
+          } else if (input.modelName.startsWith("grok")) {
+            llmClient = new AISdkClient({
+              model: wrapAISDKModel(xai(input.modelName)),
+            });
           } else if (input.modelName.includes("/")) {
             llmClient = new CustomOpenAIClient({
               modelName: input.modelName as AvailableModel,