diff --git a/src/renderer/src/pages/settings/AssistantSettings/AssistantModelSettings.tsx b/src/renderer/src/pages/settings/AssistantSettings/AssistantModelSettings.tsx
index c650011b07..c452096d47 100644
--- a/src/renderer/src/pages/settings/AssistantSettings/AssistantModelSettings.tsx
+++ b/src/renderer/src/pages/settings/AssistantSettings/AssistantModelSettings.tsx
@@ -9,6 +9,7 @@ import { DEFAULT_CONTEXTCOUNT, DEFAULT_TEMPERATURE, MAX_CONTEXT_COUNT } from '@r
 import { isEmbeddingModel, isRerankModel } from '@renderer/config/models'
 import { useTimer } from '@renderer/hooks/useTimer'
 import { SettingRow } from '@renderer/pages/settings'
+import { DEFAULT_ASSISTANT_SETTINGS } from '@renderer/services/AssistantService'
 import type { Assistant, AssistantSettingCustomParameters, AssistantSettings, Model } from '@renderer/types'
 import { modalConfirm } from '@renderer/utils'
 import { Button, Col, Divider, Input, InputNumber, Row, Select, Slider, Switch, Tooltip } from 'antd'
@@ -31,7 +32,9 @@ const AssistantModelSettings: FC<Props> = ({ assistant, updateAssistant, updateA
   const [enableMaxTokens, setEnableMaxTokens] = useState(assistant?.settings?.enableMaxTokens ?? false)
   const [maxTokens, setMaxTokens] = useState(assistant?.settings?.maxTokens ?? 0)
   const [streamOutput, setStreamOutput] = useState(assistant?.settings?.streamOutput)
-  const [toolUseMode, setToolUseMode] = useState(assistant?.settings?.toolUseMode ?? 'prompt')
+  const [toolUseMode, setToolUseMode] = useState<AssistantSettings['toolUseMode']>(
+    assistant?.settings?.toolUseMode ?? 'function'
+  )
   const [defaultModel, setDefaultModel] = useState(assistant?.defaultModel)
   const [topP, setTopP] = useState(assistant?.settings?.topP ?? 1)
   const [enableTopP, setEnableTopP] = useState(assistant?.settings?.enableTopP ?? false)
@@ -158,28 +161,17 @@ const AssistantModelSettings: FC<Props> = ({ assistant, updateAssistant, updateA
   }
 
   const onReset = () => {
-    setTemperature(DEFAULT_TEMPERATURE)
-    setEnableTemperature(true)
-    setContextCount(DEFAULT_CONTEXTCOUNT)
-    setEnableMaxTokens(false)
-    setMaxTokens(0)
-    setStreamOutput(true)
-    setTopP(1)
-    setEnableTopP(false)
-    setCustomParameters([])
-    setToolUseMode('prompt')
-    updateAssistantSettings({
-      temperature: DEFAULT_TEMPERATURE,
-      enableTemperature: true,
-      contextCount: DEFAULT_CONTEXTCOUNT,
-      enableMaxTokens: false,
-      maxTokens: 0,
-      streamOutput: true,
-      topP: 1,
-      enableTopP: false,
-      customParameters: [],
-      toolUseMode: 'prompt'
-    })
+    setTemperature(DEFAULT_ASSISTANT_SETTINGS.temperature)
+    setEnableTemperature(DEFAULT_ASSISTANT_SETTINGS.enableTemperature ?? true)
+    setContextCount(DEFAULT_ASSISTANT_SETTINGS.contextCount)
+    setEnableMaxTokens(DEFAULT_ASSISTANT_SETTINGS.enableMaxTokens ?? false)
+    setMaxTokens(DEFAULT_ASSISTANT_SETTINGS.maxTokens ?? 0)
+    setStreamOutput(DEFAULT_ASSISTANT_SETTINGS.streamOutput)
+    setTopP(DEFAULT_ASSISTANT_SETTINGS.topP)
+    setEnableTopP(DEFAULT_ASSISTANT_SETTINGS.enableTopP ?? false)
+    setCustomParameters(DEFAULT_ASSISTANT_SETTINGS.customParameters ?? [])
+    setToolUseMode(DEFAULT_ASSISTANT_SETTINGS.toolUseMode)
+    updateAssistantSettings(DEFAULT_ASSISTANT_SETTINGS)
   }
   const modelFilter = (model: Model) => !isEmbeddingModel(model) && !isRerankModel(model)
 
diff --git a/src/renderer/src/services/ApiService.ts b/src/renderer/src/services/ApiService.ts
index bc8de89403..10e191bb38 100644
--- a/src/renderer/src/services/ApiService.ts
+++ b/src/renderer/src/services/ApiService.ts
@@ -6,7 +6,7 @@ import AiProvider from '@renderer/aiCore'
 import type { CompletionsParams } from '@renderer/aiCore/legacy/middleware/schemas'
 import type { AiSdkMiddlewareConfig } from '@renderer/aiCore/middleware/AiSdkMiddlewareBuilder'
 import { buildStreamTextParams } from '@renderer/aiCore/prepareParams'
-import { isDedicatedImageGenerationModel, isEmbeddingModel } from '@renderer/config/models'
+import { isDedicatedImageGenerationModel, isEmbeddingModel, isFunctionCallingModel } from '@renderer/config/models'
 import { getStoreSetting } from '@renderer/hooks/useSettings'
 import i18n from '@renderer/i18n'
 import store from '@renderer/store'
@@ -18,6 +18,7 @@ import type { Message } from '@renderer/types/newMessage'
 import type { SdkModel } from '@renderer/types/sdk'
 import { removeSpecialCharactersForTopicName, uuid } from '@renderer/utils'
 import { abortCompletion, readyToAbort } from '@renderer/utils/abortController'
+import { isToolUseModeFunction } from '@renderer/utils/assistant'
 import { isAbortError } from '@renderer/utils/error'
 import { purifyMarkdownImages } from '@renderer/utils/markdown'
 import { isPromptToolUse, isSupportedToolUse } from '@renderer/utils/mcp-tools'
@@ -126,12 +127,16 @@ export async function fetchChatCompletion({
     requestOptions: options
   })
 
+  // Safely fallback to prompt tool use when function calling is not supported by model.
+  const usePromptToolUse =
+    isPromptToolUse(assistant) || (isToolUseModeFunction(assistant) && !isFunctionCallingModel(assistant.model))
+
   const middlewareConfig: AiSdkMiddlewareConfig = {
     streamOutput: assistant.settings?.streamOutput ?? true,
     onChunk: onChunkReceived,
     model: assistant.model,
     enableReasoning: capabilities.enableReasoning,
-    isPromptToolUse: isPromptToolUse(assistant),
+    isPromptToolUse: usePromptToolUse,
     isSupportedToolUse: isSupportedToolUse(assistant),
     isImageGenerationEndpoint: isDedicatedImageGenerationModel(assistant.model || getDefaultModel()),
     webSearchPluginConfig: webSearchPluginConfig,
diff --git a/src/renderer/src/services/AssistantService.ts b/src/renderer/src/services/AssistantService.ts
index 2aab6de23a..685ecf6324 100644
--- a/src/renderer/src/services/AssistantService.ts
+++ b/src/renderer/src/services/AssistantService.ts
@@ -36,9 +36,10 @@ export const DEFAULT_ASSISTANT_SETTINGS: AssistantSettings = {
   streamOutput: true,
   topP: 1,
   enableTopP: false,
-  toolUseMode: 'prompt',
+  // It would gracefully fallback to prompt if not supported by model.
+  toolUseMode: 'function',
   customParameters: []
-}
+} as const
 
 export function getDefaultAssistant(): Assistant {
   return {
@@ -176,7 +177,7 @@ export const getAssistantSettings = (assistant: Assistant): AssistantSettings =>
     enableMaxTokens: assistant?.settings?.enableMaxTokens ?? false,
     maxTokens: getAssistantMaxTokens(),
     streamOutput: assistant?.settings?.streamOutput ?? true,
-    toolUseMode: assistant?.settings?.toolUseMode ?? 'prompt',
+    toolUseMode: assistant?.settings?.toolUseMode ?? 'function',
     defaultModel: assistant?.defaultModel ?? undefined,
     reasoning_effort: assistant?.settings?.reasoning_effort ?? undefined,
     customParameters: assistant?.settings?.customParameters ?? []