diff --git a/CHANGELOG.md b/CHANGELOG.md
index 0a0457982c..e91bca4421 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -11,7 +11,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - Upgraded `ai` from version `7.0.114` to `7.0.127`
 
-## 3.84.0 - 2026-10-11
+## 3.84.2 - 2026-10-11
 
 ### Added
 
diff --git a/apps/api/src/app/endpoints/ai/ai.service.spec.ts b/apps/api/src/app/endpoints/ai/ai.service.spec.ts
index b2bb63addc..549bdcef35 100644
--- a/apps/api/src/app/endpoints/ai/ai.service.spec.ts
+++ b/apps/api/src/app/endpoints/ai/ai.service.spec.ts
@@ -197,7 +197,7 @@ describe('AiService', () => {
       );
     });
 
-    it('Restricts the routing to endpoints with zero data retention and passes a PDF file to the model itself', async () => {
+    it('Restricts the routing to endpoints with zero data retention, disables the reasoning and passes a PDF file to the model itself', async () => {
       await aiService.getActivityRowsFromFile({
         file: { buffer: Buffer.from('%PDF-1.7'), mediaType: 'application/pdf' },
         locale: 'de-CH'
@@ -206,6 +206,7 @@ describe('AiService', () => {
       expect(chat).toHaveBeenCalledWith('vision-model', {
         plugins: [{ id: 'file-parser', pdf: { engine: 'native' } }],
         provider: { require_parameters: true, zdr: true },
+        reasoning: { effort: 'none' },
         usage: { include: true }
       });
     });
@@ -892,7 +893,7 @@ describe('AiService', () => {
       });
     });
 
-    it('Restricts the routing to endpoints with zero data retention, limits the number of steps, requires a tool call in the first step and allows no tool call in the last step', async () => {
+    it('Restricts the routing to endpoints with zero data retention, disables the reasoning, limits the number of steps, requires a tool call in the first step and allows no tool call in the last step', async () => {
       mockRowMatches({});
 
       await aiService.getActivitiesFromRows({
@@ -905,6 +906,7 @@ describe('AiService', () => {
 
       expect(chat).toHaveBeenCalledWith('agent-model', {
         provider: { require_parameters: true, zdr: true },
+        reasoning: { effort: 'none' },
         usage: { include: true }
       });
 
diff --git a/apps/api/src/app/endpoints/ai/ai.service.ts b/apps/api/src/app/endpoints/ai/ai.service.ts
index 3c30ea8e15..006b47a358 100644
--- a/apps/api/src/app/endpoints/ai/ai.service.ts
+++ b/apps/api/src/app/endpoints/ai/ai.service.ts
@@ -289,11 +289,12 @@ export class AiService {
    * OpenRouter is not covered by the zero data retention. The model gets no
    * tools and copies the asset, the account and the broker as printed, hence
    * the rows still have to be matched to asset profiles and accounts. The
-   * name of the file is passed as well, as it can name the broker. A row above the
-   * maximum number of activities is left out. An output which reaches the
-   * maximum number of output tokens is cut, hence it gives an error for the
-   * caller. A cut output without any text is not parsed at all, hence the
-   * finish reason is checked as well.
+   * name of the file is passed as well, as it can name the broker. The
+   * reasoning is disabled, as it can add thousands of output tokens even for a
+   * single row. A row above the maximum number of activities is left out. An
+   * output which reaches the maximum number of output tokens is cut, hence it
+   * gives an error for the caller. A cut output without any text is not parsed
+   * at all, hence the finish reason is checked as well.
    */
   public async getActivityRowsFromFile({
     abortSignal,
@@ -341,6 +342,7 @@ export class AiService {
         settings: {
           plugins: [{ id: 'file-parser', pdf: { engine: 'native' } }],
           provider: { require_parameters: true, zdr: true },
+          reasoning: { effort: 'none' },
           usage: { include: true }
         }
       }),
@@ -517,10 +519,11 @@ export class AiService {
    * tools search-asset-profiles and get-accounts of the server of the Model
    * Context Protocol. The accounts are given without values. The first step
    * requires a tool call, as a model can give the matches without any tool
-   * call otherwise. The last step allows no tool call, so that the model
-   * always gives the matches. An asset match whose asset profile no search
-   * gave is left out, as the import refuses a symbol which the data source
-   * cannot resolve.
+   * call otherwise. The reasoning is disabled, as some models ignore a
+   * required tool call while they reason. The last step allows no tool call,
+   * so that the model always gives the matches. An asset match whose asset
+   * profile no search gave is left out, as the import refuses a symbol which
+   * the data source cannot resolve.
    */
   private async getRowMatches({
     abortSignal,
@@ -555,6 +558,7 @@ export class AiService {
         modelPropertyKey: PROPERTY_OPENROUTER_MODEL_AGENT,
         settings: {
           provider: { require_parameters: true, zdr: true },
+          reasoning: { effort: 'none' },
           usage: { include: true }
         }
       }),
diff --git a/package-lock.json b/package-lock.json
index 6e1efdd6a7..7522246234 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -1,12 +1,12 @@
 {
   "name": "ghostfolio",
-  "version": "3.84.0",
+  "version": "3.84.2",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "ghostfolio",
-      "version": "3.84.0",
+      "version": "3.84.2",
       "hasInstallScript": true,
       "license": "AGPL-3.0",
       "dependencies": {
diff --git a/package.json b/package.json
index ec3e7eca2d..b1879a80f2 100644
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
   "name": "ghostfolio",
-  "version": "3.84.0",
+  "version": "3.84.2",
   "homepage": "https://ghostfol.io",
   "license": "AGPL-3.0",
   "repository": "https://github.com/ghostfolio/ghostfolio",