diff --git a/src/components/ProModeSelector.tsx b/src/components/ProModeSelector.tsx
index 768ce3a..1a95a37 100644
--- a/src/components/ProModeSelector.tsx
+++ b/src/components/ProModeSelector.tsx
@@ -18,10 +18,6 @@ import { IpcClient } from "@/ipc/ipc_client";
 export function ProModeSelector() {
   const { settings, updateSettings } = useSettings();
 
-  const toggleSaverMode = () => {
-    updateSettings({ enableProSaverMode: !settings?.enableProSaverMode });
-  };
-
   const toggleLazyEdits = () => {
     updateSettings({
       enableProLazyEditsMode: !settings?.enableProLazyEditsMode,
@@ -76,16 +72,6 @@ export function ProModeSelector() {
               </a>
             </div>
           )}
-          <SelectorRow
-            id="saver-mode"
-            label="Saver Mode"
-            description="Uses your free Gemini API quota before consuming Dyad Pro AI credits"
-            tooltip="Note: using the free Gemini API lets Google use your data to
-                    improve their models."
-            proEnabled={proEnabled}
-            settingEnabled={Boolean(settings?.enableProSaverMode)}
-            toggle={toggleSaverMode}
-          />
           <SelectorRow
             id="lazy-edits"
             label="Turbo Edits"
diff --git a/src/ipc/handlers/chat_stream_handlers.ts b/src/ipc/handlers/chat_stream_handlers.ts
index ddf3aef..151c2ae 100644
--- a/src/ipc/handlers/chat_stream_handlers.ts
+++ b/src/ipc/handlers/chat_stream_handlers.ts
@@ -1,5 +1,5 @@
 import { ipcMain } from "electron";
-import { CoreMessage, TextPart, ImagePart } from "ai";
+import { CoreMessage, TextPart, ImagePart, streamText } from "ai";
 import { db } from "../../db";
 import { chats, messages } from "../../db/schema";
 import { and, eq, isNull } from "drizzle-orm";
@@ -32,7 +32,6 @@ import * as crypto from "crypto";
 import { readFile, writeFile, unlink } from "fs/promises";
 import { getMaxTokens } from "../utils/token_utils";
 import { MAX_CHAT_TURNS_IN_CONTEXT } from "@/constants/settings_constants";
-import { streamTextWithBackup } from "../utils/stream_utils";
 import { validateChatContext } from "../utils/context_paths_utils";
 
 const logger = log.scope("chat_stream_handlers");
@@ -244,8 +243,11 @@ export function registerChatStreamHandlers() {
           "estimated tokens",
           codebaseInfo.length / 4,
         );
-        const { modelClient, backupModelClients, isEngineEnabled } =
-          await getModelClient(settings.selectedModel, settings, files);
+        const { modelClient, isEngineEnabled } = await getModelClient(
+          settings.selectedModel,
+          settings,
+          files,
+        );
 
         // Prepare message history for the AI
         const messageHistory = updatedChat.messages.map((message) => ({
@@ -394,11 +396,11 @@ This conversation includes one or more image attachments. When the user uploads
         }
 
         // When calling streamText, the messages need to be properly formatted for mixed content
-        const { textStream } = streamTextWithBackup({
+        const { textStream } = streamText({
           maxTokens: await getMaxTokens(settings.selectedModel),
           temperature: 0,
-          model: modelClient,
-          backupModelClients: backupModelClients,
+          maxRetries: 2,
+          model: modelClient.model,
           system: systemPrompt,
           messages: chatMessages.filter((m) => m.content),
           onError: (error: any) => {
diff --git a/src/ipc/utils/get_model_client.ts b/src/ipc/utils/get_model_client.ts
index 693e3a0..04c99f6 100644
--- a/src/ipc/utils/get_model_client.ts
+++ b/src/ipc/utils/get_model_client.ts
@@ -10,7 +10,6 @@ import { getEnvVar } from "./read_env";
 import log from "electron-log";
 import { getLanguageModelProviders } from "../shared/language_model_helpers";
 import { LanguageModelProvider } from "../ipc_types";
-import { llmErrorStore } from "@/main/llm_error_store";
 import { createDyadEngine } from "./llm_engine_provider";
 import { findLanguageModel } from "./findLanguageModel";
 import { LM_STUDIO_BASE_URL } from "./lm_studio_utils";
@@ -50,7 +49,6 @@ export async function getModelClient(
   files?: File[],
 ): Promise<{
   modelClient: ModelClient;
-  backupModelClients: ModelClient[];
   isEngineEnabled?: boolean;
 }> {
   const allProviders = await getLanguageModelProviders();
@@ -143,42 +141,11 @@ export async function getModelClient(
         ),
         builtinProviderId: "auto",
       };
-      const googleSettings = settings.providerSettings?.google;
 
-      // Budget saver mode logic (all must be true):
-      // 1. Pro Saver Mode is enabled
-      // 2. Provider is Google
-      // 3. API Key is set
-      // 4. Has no recent errors
-      if (
-        settings.enableProSaverMode &&
-        providerConfig.id === "google" &&
-        googleSettings &&
-        googleSettings.apiKey?.value &&
-        llmErrorStore.modelHasNoRecentError({
-          model: model.name,
-          provider: providerConfig.id,
-        })
-      ) {
-        return {
-          modelClient: getRegularModelClient(
-            {
-              provider: providerConfig.id,
-              name: model.name,
-            },
-            settings,
-            providerConfig,
-          ).modelClient,
-          backupModelClients: [autoModelClient],
-          isEngineEnabled,
-        };
-      } else {
-        return {
-          modelClient: autoModelClient,
-          backupModelClients: [],
-          isEngineEnabled,
-        };
-      }
+      return {
+        modelClient: autoModelClient,
+        isEngineEnabled,
+      };
     } else {
       logger.warn(
         `Dyad Pro enabled, but provider ${model.provider} does not have a gateway prefix defined. Falling back to direct provider connection.`,
diff --git a/src/ipc/utils/stream_utils.ts b/src/ipc/utils/stream_utils.ts
deleted file mode 100644
index 59480f7..0000000
--- a/src/ipc/utils/stream_utils.ts
+++ /dev/null
@@ -1,123 +0,0 @@
-import { streamText } from "ai";
-import log from "electron-log";
-import { ModelClient } from "./get_model_client";
-import { llmErrorStore } from "@/main/llm_error_store";
-const logger = log.scope("stream_utils");
-
-export interface StreamTextWithBackupParams
-  extends Omit<Parameters<typeof streamText>[0], "model"> {
-  model: ModelClient; // primary client
-  backupModelClients?: ModelClient[]; // ordered fall-backs
-}
-
-export function streamTextWithBackup(params: StreamTextWithBackupParams): {
-  textStream: AsyncIterable<string>;
-} {
-  const {
-    model: primaryModel,
-    backupModelClients = [],
-    onError: callerOnError,
-    abortSignal: callerAbort,
-    ...rest
-  } = params;
-
-  const modelClients: ModelClient[] = [primaryModel, ...backupModelClients];
-
-  async function* combinedGenerator(): AsyncIterable<string> {
-    let lastErr: { error: unknown } | undefined = undefined;
-
-    for (let i = 0; i < modelClients.length; i++) {
-      const currentModelClient = modelClients[i];
-
-      /* Local abort controller for this single attempt  */
-      const attemptAbort = new AbortController();
-      if (callerAbort) {
-        if (callerAbort.aborted) {
-          // Already aborted, trigger immediately
-          attemptAbort.abort();
-        } else {
-          callerAbort.addEventListener("abort", () => attemptAbort.abort(), {
-            once: true,
-          });
-        }
-      }
-
-      let errorFromCurrent: { error: unknown } | undefined = undefined; // set when onError fires
-      const providerId = currentModelClient.builtinProviderId;
-      if (providerId) {
-        llmErrorStore.clearModelError({
-          model: currentModelClient.model.modelId,
-          provider: providerId,
-        });
-      }
-      logger.info(
-        "Streaming text with model",
-        currentModelClient.model.modelId,
-        "provider",
-        currentModelClient.model.provider,
-        "builtinProviderId",
-        currentModelClient.builtinProviderId,
-      );
-      const { textStream } = streamText({
-        ...rest,
-        maxRetries: i === modelClients.length - 1 ? 3 : 0,
-        model: currentModelClient.model,
-        abortSignal: attemptAbort.signal,
-        onError: (error) => {
-          const providerId = currentModelClient.builtinProviderId;
-          if (providerId) {
-            llmErrorStore.recordModelError({
-              model: currentModelClient.model.modelId,
-              provider: providerId,
-            });
-          }
-          logger.error(
-            `Error streaming text with ${providerId} and model ${currentModelClient.model.modelId}: ${error}`,
-            error,
-          );
-          errorFromCurrent = error;
-          attemptAbort.abort(); // kill fetch / SSE
-        },
-      });
-
-      try {
-        for await (const chunk of textStream) {
-          /* If onError fired during streaming, bail out immediately. */
-          if (errorFromCurrent) throw errorFromCurrent;
-          yield chunk;
-        }
-
-        /* Stream ended – check if it actually failed */
-        if (errorFromCurrent) throw errorFromCurrent;
-
-        /* Completed successfully – stop trying more models. */
-        return;
-      } catch (err) {
-        if (typeof err === "object" && err !== null && "error" in err) {
-          lastErr = err as { error: unknown };
-        } else {
-          lastErr = { error: err };
-        }
-        logger.warn(
-          `[streamTextWithBackup] model #${i} failed – ${
-            i < modelClients.length - 1
-              ? "switching to backup"
-              : "no backups left"
-          }`,
-          err,
-        );
-        /* loop continues to next model (if any) */
-      }
-    }
-
-    /* Every model failed */
-    if (!lastErr) {
-      throw new Error("Invariant in StreamTextWithbackup failed!");
-    }
-    callerOnError?.(lastErr);
-    logger.error("All model invocations failed", lastErr);
-    // throw lastErr ?? new Error("All model invocations failed");
-  }
-
-  return { textStream: combinedGenerator() };
-}
diff --git a/src/lib/schemas.ts b/src/lib/schemas.ts
index 6e478b0..904c3d5 100644
--- a/src/lib/schemas.ts
+++ b/src/lib/schemas.ts
@@ -139,7 +139,6 @@ export const UserSettingsSchema = z.object({
   experiments: ExperimentsSchema.optional(),
   lastShownReleaseNotesVersion: z.string().optional(),
   maxChatTurnsInContext: z.number().optional(),
-  enableProSaverMode: z.boolean().optional(),
   enableProLazyEditsMode: z.boolean().optional(),
   enableProSmartFilesContextMode: z.boolean().optional(),
   selectedTemplateId: z.string().optional(),
@@ -154,6 +153,7 @@ export const UserSettingsSchema = z.object({
   ////////////////////////////////
   // DEPRECATED.
   ////////////////////////////////
+  enableProSaverMode: z.boolean().optional(),
   dyadProBudget: DyadProBudgetSchema.optional(),
   runtimeMode: RuntimeModeSchema.optional(),
 });
diff --git a/src/main/llm_error_store.ts b/src/main/llm_error_store.ts
deleted file mode 100644
index 31fd4e6..0000000
--- a/src/main/llm_error_store.ts
+++ /dev/null
@@ -1,35 +0,0 @@
-class LlmErrorStore {
-  private modelErrorToTimestamp: Record<string, number> = {};
-
-  constructor() {}
-
-  recordModelError({ model, provider }: { model: string; provider: string }) {
-    this.modelErrorToTimestamp[this.getKey({ model, provider })] = Date.now();
-  }
-
-  clearModelError({ model, provider }: { model: string; provider: string }) {
-    delete this.modelErrorToTimestamp[this.getKey({ model, provider })];
-  }
-
-  modelHasNoRecentError({
-    model,
-    provider,
-  }: {
-    model: string;
-    provider: string;
-  }): boolean {
-    const key = this.getKey({ model, provider });
-    const timestamp = this.modelErrorToTimestamp[key];
-    if (!timestamp) {
-      return true;
-    }
-    const oneHourAgo = Date.now() - 1000 * 60 * 60;
-    return timestamp < oneHourAgo;
-  }
-
-  private getKey({ model, provider }: { model: string; provider: string }) {
-    return `${provider}::${model}`;
-  }
-}
-
-export const llmErrorStore = new LlmErrorStore();