diff --git a/config/llm.yml b/config/llm.yml
index 1b702ab6a..e43907234 100644
--- a/config/llm.yml
+++ b/config/llm.yml
@@ -69,6 +69,10 @@ models:
     credit_multiplier: 1
 
 features:
+  conversation_completion:
+    models: [gpt-4.1]
+    default: gpt-4.1
+    internal: true
   editor:
     models:
       [
diff --git a/enterprise/lib/captain/conversation_completion_service.rb b/enterprise/lib/captain/conversation_completion_service.rb
index 37f9add3e..47f035519 100644
--- a/enterprise/lib/captain/conversation_completion_service.rb
+++ b/enterprise/lib/captain/conversation_completion_service.rb
@@ -16,7 +16,8 @@ class Captain::ConversationCompletionService < Captain::BaseTaskService
     return default_incomplete_response('No messages found') if content.blank?
 
     response = make_api_call(
-      model: InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_MODEL')&.value.presence || GPT_MODEL,
+      feature: 'conversation_completion',
+      model: self_hosted_model_override,
       messages: [
         { role: 'system', content: prompt_from_file('conversation_completion') },
         { role: 'user', content: content }
@@ -31,6 +32,12 @@ class Captain::ConversationCompletionService < Captain::BaseTaskService
 
   private
 
+  def self_hosted_model_override
+    return unless ChatwootApp.self_hosted_enterprise?
+
+    InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_MODEL')&.value.presence
+  end
+
   def prompt_from_file(file_name)
     Rails.root.join('enterprise/lib/captain/prompts', "#{file_name}.liquid").read
   end
diff --git a/lib/llm/feature_router.rb b/lib/llm/feature_router.rb
index d75b06dca..cb5a1f256 100644
--- a/lib/llm/feature_router.rb
+++ b/lib/llm/feature_router.rb
@@ -24,6 +24,8 @@ module Llm::FeatureRouter
     private
 
     def account_model_override(account, feature_key)
+      return if Llm::Models.internal_feature?(feature_key)
+
       model = account&.captain_models&.[](feature_key).presence
       return unless model
       return model if Llm::Models.valid_model_for?(feature_key, model)
diff --git a/lib/llm/models.rb b/lib/llm/models.rb
index 896014262..63e8bdfe4 100644
--- a/lib/llm/models.rb
+++ b/lib/llm/models.rb
@@ -5,7 +5,11 @@ module Llm::Models
     def providers = CONFIG.fetch('providers')
     def models = CONFIG.fetch('models')
     def features = CONFIG.fetch('features')
-    def feature_keys = features.keys
+    def feature_keys = features.reject { |_key, config| config['internal'] }.keys
+
+    def internal_feature?(feature)
+      features.dig(feature.to_s, 'internal') == true
+    end
 
     def feature?(feature)
       features.key?(feature.to_s)
diff --git a/spec/controllers/api/v1/accounts/captain/preferences_controller_spec.rb b/spec/controllers/api/v1/accounts/captain/preferences_controller_spec.rb
index 6a28c60da..016d7d4d5 100644
--- a/spec/controllers/api/v1/accounts/captain/preferences_controller_spec.rb
+++ b/spec/controllers/api/v1/accounts/captain/preferences_controller_spec.rb
@@ -31,6 +31,7 @@ RSpec.describe 'Api::V1::Accounts::Captain::Preferences', type: :request do
         expect(json_response).to have_key(:providers)
         expect(json_response).to have_key(:models)
         expect(json_response).to have_key(:features)
+        expect(json_response[:features]).not_to have_key(:conversation_completion)
       end
     end
 
@@ -150,14 +151,21 @@ RSpec.describe 'Api::V1::Accounts::Captain::Preferences', type: :request do
         expect(account.reload.captain_models['editor']).to eq('gpt-4.1-mini')
       end
 
-      it 'does not persist unknown captain model feature keys' do
+      it 'does not persist unknown or internal captain model feature keys' do
         put "/api/v1/accounts/#{account.id}/captain/preferences",
             headers: admin.create_new_auth_token,
-            params: { captain_models: { editor: 'gpt-4.1-mini', unknown_feature: 'gpt-4.1' } },
+            params: {
+              captain_models: {
+                editor: 'gpt-4.1-mini',
+                unknown_feature: 'gpt-4.1',
+                conversation_completion: 'gpt-4.1'
+              }
+            },
             as: :json
 
         expect(response).to have_http_status(:success)
         expect(account.reload.captain_models).to eq('editor' => 'gpt-4.1-mini')
+        expect(json_response[:features]).not_to have_key(:conversation_completion)
       end
 
       it 'rejects invalid captain model values for the feature' do
diff --git a/spec/controllers/super_admin/accounts_controller_spec.rb b/spec/controllers/super_admin/accounts_controller_spec.rb
index 29f93c914..6da2dbc2f 100644
--- a/spec/controllers/super_admin/accounts_controller_spec.rb
+++ b/spec/controllers/super_admin/accounts_controller_spec.rb
@@ -57,6 +57,7 @@ RSpec.describe 'Super Admin accounts API', type: :request do
         Llm::Models.feature_keys.each do |feature_key|
           expect(response.body).to include("account[captain_models][#{feature_key}]")
         end
+        expect(response.body).not_to include('account[captain_models][conversation_completion]')
 
         document = Nokogiri::HTML(response.body)
         editor_select = document.at_css('select[name="account[captain_models][editor]"]')
diff --git a/spec/enterprise/lib/captain/conversation_completion_service_spec.rb b/spec/enterprise/lib/captain/conversation_completion_service_spec.rb
index 525f2954e..b4d787132 100644
--- a/spec/enterprise/lib/captain/conversation_completion_service_spec.rb
+++ b/spec/enterprise/lib/captain/conversation_completion_service_spec.rb
@@ -19,6 +19,42 @@ RSpec.describe Captain::ConversationCompletionService do
   end
 
   describe '#perform' do
+    describe 'model routing' do
+      let(:mock_response) do
+        instance_double(RubyLLM::Message, content: { 'complete' => true, 'reason' => 'Done' }, input_tokens: 10, output_tokens: 5)
+      end
+
+      before do
+        create(:message, conversation: conversation, message_type: :incoming, content: 'Hello')
+        allow(mock_chat).to receive(:ask).and_return(mock_response)
+      end
+
+      it 'uses the internal GPT-4.1 route on Chatwoot Cloud' do
+        allow(ChatwootApp).to receive(:self_hosted_enterprise?).and_return(false)
+        InstallationConfig.find_or_initialize_by(name: 'CAPTAIN_OPEN_AI_MODEL').update!(value: 'gpt-5.1')
+        account.enable_features!('captain_integration_v2')
+        allow(mock_context).to receive(:chat).with(model: 'gpt-4.1').and_return(mock_chat)
+
+        expect(service.perform).to include(complete: true)
+      end
+
+      it 'uses the installation model on self-hosted Enterprise' do
+        allow(ChatwootApp).to receive(:self_hosted_enterprise?).and_return(true)
+        InstallationConfig.find_or_initialize_by(name: 'CAPTAIN_OPEN_AI_MODEL').update!(value: 'gpt-5.1')
+        allow(mock_context).to receive(:chat).with(model: 'gpt-5.1').and_return(mock_chat)
+
+        expect(service.perform).to include(complete: true)
+      end
+
+      it 'falls back to the internal GPT-4.1 route when the self-hosted installation model is blank' do
+        allow(ChatwootApp).to receive(:self_hosted_enterprise?).and_return(true)
+        InstallationConfig.find_or_initialize_by(name: 'CAPTAIN_OPEN_AI_MODEL').update!(value: '')
+        allow(mock_context).to receive(:chat).with(model: 'gpt-4.1').and_return(mock_chat)
+
+        expect(service.perform).to include(complete: true)
+      end
+    end
+
     context 'when conversation is complete' do
       let(:mock_response) do
         instance_double(