diff --git a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatUI.tsx b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatUI.tsx
index 26585d45829..5cce2bd04cf 100644
--- a/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatUI.tsx
+++ b/ui/litellm-dashboard/src/app/(dashboard)/playground/components/chat_ui/ChatUI.tsx
@@ -455,7 +455,7 @@ const ChatUI: React.FC<ChatUIProps> = ({
     if (!simplified) {
       void loadModels();
     }
-    void loadMCPServers(userApiKey);
+    void loadMCPServers();
 
     return () => {
       cancelled = true;
@@ -794,7 +794,8 @@ const ChatUI: React.FC<ChatUIProps> = ({
       return;
     }
 
-    const effectiveApiKey = simplified ? accessToken : apiKeySource === "session" ? accessToken : apiKey;
+    const effectiveApiKey =
+      simplified || apiKeySource === "session" ? accessToken : debouncedCustomApiKey || apiKey.trim();
 
     if (!effectiveApiKey) {
       NotificationsManager.fromBackend("Please provide a Virtual Key or select Current UI Session");
diff --git a/ui/litellm-dashboard/src/components/llm_calls/fetch_models.test.ts b/ui/litellm-dashboard/src/components/llm_calls/fetch_models.test.ts
index f08143b52f2..d00c17b306a 100644
--- a/ui/litellm-dashboard/src/components/llm_calls/fetch_models.test.ts
+++ b/ui/litellm-dashboard/src/components/llm_calls/fetch_models.test.ts
@@ -74,4 +74,12 @@ describe("fetchAvailableModels", () => {
     const models = await fetchAvailableModels("sk-virtual-key");
     expect(models).toEqual([{ model_group: "only-from-key", mode: undefined }]);
   });
+
+  it("throws when both model endpoints fail", async () => {
+    mockGet.mockImplementation(async () => {
+      throw new Error("network down");
+    });
+
+    await expect(fetchAvailableModels("sk-virtual-key")).rejects.toThrow("network down");
+  });
 });
diff --git a/ui/litellm-dashboard/src/components/llm_calls/fetch_models.tsx b/ui/litellm-dashboard/src/components/llm_calls/fetch_models.tsx
index cc74355539f..d1e90fda43d 100644
--- a/ui/litellm-dashboard/src/components/llm_calls/fetch_models.tsx
+++ b/ui/litellm-dashboard/src/components/llm_calls/fetch_models.tsx
@@ -25,15 +25,10 @@ const dedupeAndSort = (models: ModelGroup[]): ModelGroup[] => {
   return unique;
 };
 
-/**
- * Loads models available to the given key.
- *
- * Prefers OpenAI-compatible `/v1/models` (scoped to the key's access) and enriches
- * entries with `mode` from `/model_group/info` when that endpoint is available.
- * Falls back to `/model_group/info` alone if `/v1/models` is empty or fails.
- */
 export const fetchAvailableModels = async (accessToken: string): Promise<ModelGroup[]> => {
   const modeByName = new Map<string, string | undefined>();
+  let groupInfoError: unknown;
+  let listModelsError: unknown;
 
   try {
     const groupInfo = await apiClient.get<{ data?: ModelGroupInfoItem[] }>("/model_group/info", {
@@ -46,6 +41,7 @@ export const fetchAvailableModels = async (accessToken: string): Promise<ModelGr
       }
     }
   } catch (error) {
+    groupInfoError = error;
     console.error("Error fetching model group info:", error);
   }
 
@@ -70,6 +66,7 @@ export const fetchAvailableModels = async (accessToken: string): Promise<ModelGr
       return dedupeAndSort(fromList);
     }
   } catch (error) {
+    listModelsError = error;
     console.error("Error fetching /v1/models:", error);
   }
 
@@ -82,5 +79,11 @@ export const fetchAvailableModels = async (accessToken: string): Promise<ModelGr
     );
   }
 
+  if (groupInfoError && listModelsError) {
+    throw listModelsError instanceof Error
+      ? listModelsError
+      : new Error("Failed to load models from /v1/models and /model_group/info");
+  }
+
   return [];
 };