diff --git a/web/containers/Providers/DataLoader.tsx b/web/containers/Providers/DataLoader.tsx
index afa47b42eb..1f2b8c835d 100644
--- a/web/containers/Providers/DataLoader.tsx
+++ b/web/containers/Providers/DataLoader.tsx
@@ -1,11 +1,21 @@
 'use client'
 
-import { useEffect } from 'react'
+import { useEffect, useMemo } from 'react'
 
-import { useAtomValue } from 'jotai'
+import { Engine } from '@cortexso/cortex.js/resources'
+import {
+  EngineStatus,
+  LocalEngine,
+  LocalEngines,
+  Model,
+  RemoteEngine,
+  RemoteEngines,
+} from '@janhq/core'
+import { useAtomValue, useSetAtom } from 'jotai'
 
 import useAssistantCreate, { janAssistant } from '@/hooks/useAssistantCreate'
 import useAssistantQuery from '@/hooks/useAssistantQuery'
+import useCortex from '@/hooks/useCortex'
 import useEngineQuery from '@/hooks/useEngineQuery'
 import { useLoadTheme } from '@/hooks/useLoadTheme'
 import useModelHub from '@/hooks/useModelHub'
@@ -13,17 +23,24 @@ import useModelQuery from '@/hooks/useModelQuery'
 import useThreadCreateMutation from '@/hooks/useThreadCreateMutation'
 import useThreadQuery from '@/hooks/useThreadQuery'
 
-import { getSelectedModelAtom } from '@/helpers/atoms/Model.atom'
+import {
+  getSelectedModelAtom,
+  updateSelectedModelAtom,
+} from '@/helpers/atoms/Model.atom'
 import { threadsAtom } from '@/helpers/atoms/Thread.atom'
 
 const DataLoader: React.FC = () => {
   const selectedModel = useAtomValue(getSelectedModelAtom)
+  const setSelectedModel = useSetAtom(updateSelectedModelAtom)
   const allThreads = useAtomValue(threadsAtom)
   const { data: assistants } = useAssistantQuery()
   const { data: models } = useModelQuery()
   const { data: threads, isLoading: isFetchingThread } = useThreadQuery()
+  const { data: engineData } = useEngineQuery()
+  const { data: modelHubData } = useModelHub()
   const createThreadMutation = useThreadCreateMutation()
   const assistantCreateMutation = useAssistantCreate()
+  const { createModel } = useCortex()
 
   useEffect(() => {
     if (!assistants) return
@@ -34,19 +51,105 @@ const DataLoader: React.FC = () => {
     }
   }, [assistants, assistantCreateMutation])
 
+  const isAnyRemoteModelConfigured = useMemo(() => {
+    if (!engineData) return false
+
+    let result = false
+    for (const engine of engineData) {
+      if (RemoteEngines.includes(engine.name as RemoteEngine)) {
+        if (engine.status === EngineStatus.Ready) {
+          result = true
+        }
+      }
+    }
+    return result
+  }, [engineData])
+
+  const isAnyModelReady = useMemo(() => {
+    if (!models) return false
+    return models.length > 0
+  }, [models])
+
   // automatically create new thread if thread list is empty
   useEffect(() => {
     if (isFetchingThread) return
     if (allThreads.length > 0) return
     if (!assistants || assistants.length === 0) return
-    if (!models || models.length === 0) return
-    if (allThreads.length === 0 && !createThreadMutation.isPending) {
-      const model = selectedModel ?? models[0]
-      const assistant = assistants[0]
+    const shouldCreateNewThread = isAnyRemoteModelConfigured || isAnyModelReady
+
+    if (shouldCreateNewThread && !createThreadMutation.isPending) {
+      // if we already have selected model then can safely proceed
+      if (selectedModel) {
+        const assistant = assistants[0]
 
-      console.log('Create new thread because user have no thread')
+        console.debug(
+          'Create new thread because user have no thread, with selected model',
+          selectedModel.model
+        )
+        createThreadMutation.mutate({
+          modelId: selectedModel.model,
+          assistant: assistant,
+        })
+        return
+      }
+
+      let modelToBeUsed: Model | undefined = undefined
+      // if we have a model registered already, try to use it and prioritize local model
+      if (models && models.length > 0) {
+        for (const model of models) {
+          if (!model.engine) continue
+          if (LocalEngines.includes(model.engine as LocalEngine)) {
+            modelToBeUsed = model
+          }
+        }
+
+        // if we don't have it, then just take the first one
+        if (!modelToBeUsed) {
+          modelToBeUsed = models[0]
+        }
+      } else {
+        if (!engineData) return
+        // we don't have nay registered model, so will need to check the remote engine
+        const remoteEngineReadyList: Engine[] = []
+        for (const engine of engineData) {
+          if (RemoteEngines.includes(engine.name as RemoteEngine)) {
+            if (engine.status === EngineStatus.Ready) {
+              remoteEngineReadyList.push(engine)
+            }
+          }
+        }
+
+        if (remoteEngineReadyList.length === 0) {
+          console.debug("No remote engine ready, can't create thread")
+          return
+        }
+        // find the model from hub that using the engine
+        if (!modelHubData) return
+        const remoteEngineReadyNames = remoteEngineReadyList.map((e) => e.name)
+
+        console.log('remoteEngineReady:', remoteEngineReadyNames)
+        // loop through the modelHubData.modelCategories to find the model that using the engine
+        for (const [key, value] of modelHubData.modelCategories) {
+          if (remoteEngineReadyNames.includes(key) && value.length > 0) {
+            modelToBeUsed = value[0].model
+            if (modelToBeUsed) break
+          }
+        }
+      }
+
+      if (!modelToBeUsed) {
+        console.debug('No model to be used')
+        return
+      }
+      console.log(
+        'Create new thread because user have no thread, model to be used:',
+        modelToBeUsed.model
+      )
+      createModel(modelToBeUsed)
+      setSelectedModel(modelToBeUsed)
+      const assistant = assistants[0]
       createThreadMutation.mutate({
-        modelId: model.id,
+        modelId: modelToBeUsed.model,
         assistant: assistant,
       })
     }
@@ -58,11 +161,15 @@ const DataLoader: React.FC = () => {
     createThreadMutation,
     allThreads,
     selectedModel,
+    isAnyModelReady,
+    isAnyRemoteModelConfigured,
+    engineData,
+    modelHubData,
+    setSelectedModel,
+    createModel,
   ])
 
-  useModelHub()
   useLoadTheme()
-  useEngineQuery()
 
   return null
 }
diff --git a/web/containers/Providers/DownloadEventListener.tsx b/web/containers/Providers/DownloadEventListener.tsx
index 9ec6382ae7..bb60fa41dc 100644
--- a/web/containers/Providers/DownloadEventListener.tsx
+++ b/web/containers/Providers/DownloadEventListener.tsx
@@ -2,11 +2,12 @@ import { useCallback, useEffect, useRef } from 'react'
 
 import { DownloadState2 } from '@janhq/core'
 import { fetchEventSource } from '@microsoft/fetch-event-source'
+import { useQueryClient } from '@tanstack/react-query'
 import { useAtomValue, useSetAtom } from 'jotai'
 
 import { downloadStateListAtom } from '@/hooks/useDownloadState'
 
-import useModels from '@/hooks/useModels'
+import { modelQueryKey } from '@/hooks/useModelQuery'
 
 import { waitingForCortexAtom } from '@/helpers/atoms/App.atom'
 import { hostAtom } from '@/helpers/atoms/AppConfig.atom'
@@ -21,12 +22,12 @@ const DownloadEventListener: React.FC = () => {
   const abortController = useRef(new AbortController())
   const setDownloadStateList = useSetAtom(downloadStateListAtom)
   const setWaitingForCortex = useSetAtom(waitingForCortexAtom)
-  const { getModels } = useModels()
 
   const updateImportingModelProgress = useSetAtom(
     updateImportingModelProgressAtom
   )
   const setImportingModelSuccess = useSetAtom(setImportingModelSuccessAtom)
+  const queryClient = useQueryClient()
 
   const handleLocalImportModels = useCallback(
     (events: DownloadState2[]) => {
@@ -38,9 +39,10 @@ const DownloadEventListener: React.FC = () => {
           updateImportingModelProgress(event.id, event.progress)
         }
       }
-      getModels()
+
+      queryClient.invalidateQueries({ queryKey: modelQueryKey })
     },
-    [setImportingModelSuccess, updateImportingModelProgress, getModels]
+    [setImportingModelSuccess, updateImportingModelProgress, queryClient]
   )
 
   const subscribeDownloadEvent = useCallback(async () => {
@@ -54,7 +56,6 @@ const DownloadEventListener: React.FC = () => {
           const localImportEvents: DownloadState2[] = []
           // filter out the import local events
           for (const event of downloadEvents) {
-            console.debug('Receiving event', event)
             if (
               isAbsolutePath(event.id) &&
               event.type === 'model' &&
diff --git a/web/containers/Providers/ModalMigrations.tsx b/web/containers/Providers/ModalMigrations.tsx
index 6f3cc6abc1..8f1746674d 100644
--- a/web/containers/Providers/ModalMigrations.tsx
+++ b/web/containers/Providers/ModalMigrations.tsx
@@ -2,6 +2,7 @@ import React, { Fragment, useCallback, useMemo, useState } from 'react'
 
 import { Button, Modal, Badge } from '@janhq/joi'
 
+import { useQueryClient } from '@tanstack/react-query'
 import { atom, useAtom, useSetAtom } from 'jotai'
 import { AlertTriangleIcon } from 'lucide-react'
 
@@ -11,7 +12,7 @@ import Spinner from '@/containers/Loader/Spinner'
 
 import useMigratingData from '@/hooks/useMigratingData'
 
-import useModels from '@/hooks/useModels'
+import { modelQueryKey } from '@/hooks/useModelQuery'
 
 import { didShowMigrationWarningAtom } from '@/helpers/atoms/AppConfig.atom'
 
@@ -31,7 +32,7 @@ const ModalMigrations = () => {
     useState<MigrationState>('idle')
   const [modelMigrationState, setModelMigrationState] =
     useState<MigrationState>('idle')
-  const { getModels } = useModels()
+  const queryClient = useQueryClient()
 
   const getStepTitle = () => {
     switch (step) {
@@ -74,8 +75,8 @@ const ModalMigrations = () => {
     setStep(2)
     await migratingModels()
     await migrationThreadsAndMessages()
-    getModels()
-  }, [migratingModels, migrationThreadsAndMessages, getModels])
+    queryClient.invalidateQueries({ queryKey: modelQueryKey })
+  }, [migratingModels, migrationThreadsAndMessages, queryClient])
 
   const onDismiss = useCallback(() => {
     setStep(1)
diff --git a/web/containers/Providers/ModelEventListener.tsx b/web/containers/Providers/ModelEventListener.tsx
index 110f80c692..b872898d16 100644
--- a/web/containers/Providers/ModelEventListener.tsx
+++ b/web/containers/Providers/ModelEventListener.tsx
@@ -7,10 +7,11 @@ import {
   StatusAndEvent,
 } from '@janhq/core'
 import { fetchEventSource } from '@microsoft/fetch-event-source'
+import { useQueryClient } from '@tanstack/react-query'
 import { useAtomValue, useSetAtom } from 'jotai'
 
 import { removeDownloadSuccessItemAtom } from '@/hooks/useDownloadState'
-import useModels from '@/hooks/useModels'
+import { modelQueryKey } from '@/hooks/useModelQuery'
 
 import { toaster } from '../Toast'
 
@@ -25,7 +26,7 @@ function ModelEventListener() {
   const removeDownloadSuccessItem = useSetAtom(removeDownloadSuccessItemAtom)
   const setIsLoadingModel = useSetAtom(isLoadingModelAtom)
 
-  const { getModels } = useModels()
+  const queryClient = useQueryClient()
 
   const handleModelEvent = useCallback(
     (modelEvent: ModelEvent) => {
@@ -64,11 +65,11 @@ function ModelEventListener() {
 
         case 'model-downloaded':
           removeDownloadSuccessItem(modelEvent.model)
-          getModels()
+          queryClient.invalidateQueries({ queryKey: modelQueryKey })
           break
 
         case 'model-deleted':
-          getModels()
+          queryClient.invalidateQueries({ queryKey: modelQueryKey })
           break
 
         case 'stopping-failed':
@@ -84,7 +85,7 @@ function ModelEventListener() {
           break
       }
     },
-    [getModels, removeDownloadSuccessItem, setIsLoadingModel]
+    [removeDownloadSuccessItem, setIsLoadingModel, queryClient]
   )
 
   const subscribeModelEvent = useCallback(async () => {
diff --git a/web/helpers/atoms/SetupRemoteModel.atom.ts b/web/helpers/atoms/SetupRemoteModel.atom.ts
index 798e6d02fd..a3a07eda8c 100644
--- a/web/helpers/atoms/SetupRemoteModel.atom.ts
+++ b/web/helpers/atoms/SetupRemoteModel.atom.ts
@@ -1,14 +1,8 @@
 import { RemoteEngine } from '@janhq/core'
 import { atom } from 'jotai'
-import { atomWithStorage } from 'jotai/utils'
 
 export type SetupRemoteModelStage = 'NONE' | 'SETUP_INTRO' | 'SETUP_API_KEY'
-const IS_ANY_REMOTE_MODEL_CONFIGURED = 'isAnyRemoteModelConfigured'
 
-export const isAnyRemoteModelConfiguredAtom = atomWithStorage(
-  IS_ANY_REMOTE_MODEL_CONFIGURED,
-  false
-)
 const remoteModelSetUpStageAtom = atom<SetupRemoteModelStage>('NONE')
 const engineBeingSetUpAtom = atom<RemoteEngine | undefined>(undefined)
 const remoteEngineBeingSetUpMetadataAtom = atom<
diff --git a/web/hooks/useModelHub.ts b/web/hooks/useModelHub.ts
index 7b29f70163..a87de67c8d 100644
--- a/web/hooks/useModelHub.ts
+++ b/web/hooks/useModelHub.ts
@@ -168,11 +168,14 @@ const useModelHub = () => {
         results[1].data.forEach((modelEntry) => {
           const engine = modelEntry.engine
           if (modelEntry.remoteModel === true && engine) {
-            // @ts-expect-error ignore
-            data.modelCategories[engine] = data.modelCategories[engine]
-              ? // @ts-expect-error ignore
-                [...data.modelCategories[engine], modelEntry]
-              : [modelEntry]
+            if (data.modelCategories.has(engine)) {
+              data.modelCategories.set(
+                engine,
+                data.modelCategories.get(engine)!.concat(modelEntry)
+              )
+            } else {
+              data.modelCategories.set(engine, [modelEntry])
+            }
           }
         })
       }
diff --git a/web/hooks/useModels.ts b/web/hooks/useModels.ts
index 40a42da413..4d097a6da1 100644
--- a/web/hooks/useModels.ts
+++ b/web/hooks/useModels.ts
@@ -6,27 +6,12 @@ import { toaster } from '@/containers/Toast'
 
 import useCortex from './useCortex'
 
-import {
-  downloadedModelsAtom,
-  removeDownloadedModelAtom,
-} from '@/helpers/atoms/Model.atom'
+import { removeDownloadedModelAtom } from '@/helpers/atoms/Model.atom'
 
 const useModels = () => {
-  const setDownloadedModels = useSetAtom(downloadedModelsAtom)
   const removeDownloadedModel = useSetAtom(removeDownloadedModelAtom)
-  const {
-    fetchModels,
-    deleteModel: cortexDeleteModel,
-    updateModel: cortexUpdateModel,
-  } = useCortex()
-
-  const getModels = useCallback(() => {
-    const getDownloadedModels = async () => {
-      const models = await fetchModels()
-      setDownloadedModels(models)
-    }
-    getDownloadedModels()
-  }, [setDownloadedModels, fetchModels])
+  const { deleteModel: cortexDeleteModel, updateModel: cortexUpdateModel } =
+    useCortex()
 
   const deleteModel = useCallback(
     async (modelId: string) => {
@@ -48,7 +33,7 @@ const useModels = () => {
     [cortexUpdateModel]
   )
 
-  return { getModels, deleteModel, updateModel }
+  return { deleteModel, updateModel }
 }
 
 export default useModels
diff --git a/web/screens/HubScreen2/components/HubModelCard.tsx b/web/screens/HubScreen2/components/HubModelCard.tsx
index aa6d934ad9..667bc56c1c 100644
--- a/web/screens/HubScreen2/components/HubModelCard.tsx
+++ b/web/screens/HubScreen2/components/HubModelCard.tsx
@@ -3,6 +3,7 @@ import React, { useCallback, useMemo } from 'react'
 import { EngineStatus, LocalEngines, RemoteEngine } from '@janhq/core'
 
 import { Button } from '@janhq/joi'
+import { useQueryClient } from '@tanstack/react-query'
 import { useAtomValue, useSetAtom } from 'jotai'
 
 import { CloudDownload } from 'lucide-react'
@@ -14,7 +15,7 @@ import useAssistantQuery from '@/hooks/useAssistantQuery'
 import useCortex from '@/hooks/useCortex'
 
 import useEngineQuery from '@/hooks/useEngineQuery'
-import useModels from '@/hooks/useModels'
+import { modelQueryKey } from '@/hooks/useModelQuery'
 import useThreads from '@/hooks/useThreads'
 
 import { HfModelEntry } from '@/utils/huggingface'
@@ -32,6 +33,7 @@ const HubModelCard: React.FC<HfModelEntry> = ({ name, downloads, model }) => {
   const downloadedModels = useAtomValue(downloadedModelsAtom)
   const { data: assistants } = useAssistantQuery()
   const { data: engineData } = useEngineQuery()
+  const queryClient = useQueryClient()
 
   const setUpRemoteModelStage = useSetAtom(setUpRemoteModelStageAtom)
   const setLocalModelModalStage = useSetAtom(localModelModalStageAtom)
@@ -39,7 +41,6 @@ const HubModelCard: React.FC<HfModelEntry> = ({ name, downloads, model }) => {
   const { createThread } = useThreads()
   const setMainViewState = useSetAtom(mainViewStateAtom)
   const { createModel } = useCortex()
-  const { getModels } = useModels()
 
   const isLocalModel = useMemo(
     () =>
@@ -131,13 +132,12 @@ const HubModelCard: React.FC<HfModelEntry> = ({ name, downloads, model }) => {
 
       if (isApiKeyAdded) {
         createModel(model).then(() => {
-          getModels()
+          queryClient.invalidateQueries({ queryKey: modelQueryKey })
         })
         return
       }
     }
   }, [
-    getModels,
     createModel,
     createThread,
     setMainViewState,
@@ -149,6 +149,7 @@ const HubModelCard: React.FC<HfModelEntry> = ({ name, downloads, model }) => {
     isLocalModel,
     downloadedModels,
     assistants,
+    queryClient,
   ])
 
   const owner = model?.metadata?.owned_by ?? ''
diff --git a/web/screens/HubScreen2/components/RemoteModelCard.tsx b/web/screens/HubScreen2/components/RemoteModelCard.tsx
index b454f03d98..0d2978307c 100644
--- a/web/screens/HubScreen2/components/RemoteModelCard.tsx
+++ b/web/screens/HubScreen2/components/RemoteModelCard.tsx
@@ -1,6 +1,8 @@
 import React, { useCallback } from 'react'
 
 import { EngineStatus, RemoteEngine } from '@janhq/core'
+import { useQueryClient } from '@tanstack/react-query'
+
 import { useAtomValue, useSetAtom } from 'jotai'
 
 import { toaster } from '@/containers/Toast'
@@ -10,7 +12,7 @@ import useAssistantQuery from '@/hooks/useAssistantQuery'
 import useCortex from '@/hooks/useCortex'
 
 import useEngineQuery from '@/hooks/useEngineQuery'
-import useModels from '@/hooks/useModels'
+import { modelQueryKey } from '@/hooks/useModelQuery'
 import useThreads from '@/hooks/useThreads'
 
 import { HfModelEntry } from '@/utils/huggingface'
@@ -23,12 +25,11 @@ const RemoteModelCard: React.FC<HfModelEntry> = ({ name, model }) => {
   const { createThread } = useThreads()
   const setMainViewState = useSetAtom(mainViewStateAtom)
   const setUpRemoteModelStage = useSetAtom(setUpRemoteModelStageAtom)
+  const queryClient = useQueryClient()
 
   const { createModel } = useCortex()
-  const { getModels } = useModels()
   const downloadedModels = useAtomValue(downloadedModelsAtom)
   const { data: assistants } = useAssistantQuery()
-
   const { data: engineData } = useEngineQuery()
 
   const modelDisplayName = model?.name ?? name
@@ -85,7 +86,7 @@ const RemoteModelCard: React.FC<HfModelEntry> = ({ name, model }) => {
     if (isApiKeyAdded) {
       // TODO: useMutation reactQuery?
       await createModel(model)
-      getModels()
+      queryClient.invalidateQueries({ queryKey: modelQueryKey })
       if (!assistants || assistants.length === 0) {
         toaster({
           title: 'No assistant available.',
@@ -109,11 +110,11 @@ const RemoteModelCard: React.FC<HfModelEntry> = ({ name, model }) => {
     createModel,
     createThread,
     downloadedModels,
-    getModels,
     model,
     setMainViewState,
     setUpRemoteModelStage,
     modelDisplayName,
+    queryClient,
   ])
 
   return (
diff --git a/web/screens/HubScreen2/components/SetUpApiKeyModal.tsx b/web/screens/HubScreen2/components/SetUpApiKeyModal.tsx
index 67b83555e7..05fb8a9fd9 100644
--- a/web/screens/HubScreen2/components/SetUpApiKeyModal.tsx
+++ b/web/screens/HubScreen2/components/SetUpApiKeyModal.tsx
@@ -4,7 +4,7 @@ import Image from 'next/image'
 
 import { EngineStatus } from '@janhq/core'
 import { Button, Input, Modal } from '@janhq/joi'
-import { useAtom, useSetAtom } from 'jotai'
+import { useAtom } from 'jotai'
 import { ArrowUpRight } from 'lucide-react'
 
 import useEngineMutation from '@/hooks/useEngineMutation'
@@ -12,13 +12,10 @@ import useEngineQuery from '@/hooks/useEngineQuery'
 
 import { getTitleByCategory } from '@/utils/model-engine'
 
-import { isAnyRemoteModelConfiguredAtom } from '@/helpers/atoms/SetupRemoteModel.atom'
-
 import { setUpRemoteModelStageAtom } from '@/helpers/atoms/SetupRemoteModel.atom'
 
 const SetUpApiKeyModal: React.FC = () => {
   const updateEngineConfig = useEngineMutation()
-  const isAnyRemoteModelConfigured = useSetAtom(isAnyRemoteModelConfiguredAtom)
   const { data: engineData } = useEngineQuery()
 
   const [{ stage, remoteEngine, metadata }, setUpRemoteModelStage] = useAtom(
@@ -50,8 +47,7 @@ const SetUpApiKeyModal: React.FC = () => {
         value: apiKey,
       },
     })
-    isAnyRemoteModelConfigured(true)
-  }, [remoteEngine, updateEngineConfig, apiKey, isAnyRemoteModelConfigured])
+  }, [remoteEngine, updateEngineConfig, apiKey])
 
   const onDismiss = useCallback(() => {
     setUpRemoteModelStage('NONE', undefined)
diff --git a/web/screens/HubScreen2/index.tsx b/web/screens/HubScreen2/index.tsx
index d800df2e9c..b5e5f64efe 100644
--- a/web/screens/HubScreen2/index.tsx
+++ b/web/screens/HubScreen2/index.tsx
@@ -44,11 +44,11 @@ const HubScreen2: React.FC = () => {
   if (!data) return <div>Failed to fetch models</div>
 
   const engineModelMap = new Map<typeof RemoteEngines, HfModelEntry[]>()
-  Object.entries(data.modelCategories).forEach(([key, value]) => {
+  for (const [key, value] of data.modelCategories) {
     if (key !== 'HuggingFace' && key !== 'BuiltInModels') {
       engineModelMap.set(key as unknown as typeof RemoteEngines, value)
     }
-  })
+  }
 
   if (detailCategory) {
     return (
diff --git a/web/screens/Settings/MyModels/ModelItem/index.tsx b/web/screens/Settings/MyModels/ModelItem/index.tsx
index 4c2519e86c..5314ba96ed 100644
--- a/web/screens/Settings/MyModels/ModelItem/index.tsx
+++ b/web/screens/Settings/MyModels/ModelItem/index.tsx
@@ -21,6 +21,7 @@ import { twMerge } from 'tailwind-merge'
 import { toaster } from '@/containers/Toast'
 
 import useAssistantQuery from '@/hooks/useAssistantQuery'
+import useEngineInit from '@/hooks/useEngineInit'
 import useEngineQuery from '@/hooks/useEngineQuery'
 import useModelStart from '@/hooks/useModelStart'
 import useModelStop from '@/hooks/useModelStop'
@@ -54,6 +55,7 @@ const ModelItem: React.FC<Props> = ({ model }) => {
   const isEngineReady =
     engineData?.find((e) => e.name === model.engine)?.status ===
     EngineStatus.Ready
+  const initializeEngine = useEngineInit()
 
   const [menu, setMenu] = useState<HTMLDivElement | null>(null)
   const [toggle, setToggle] = useState<HTMLDivElement | null>(null)
@@ -74,6 +76,46 @@ const ModelItem: React.FC<Props> = ({ model }) => {
         stopModel.mutate(modelId)
         return
       }
+      const modelEngine = model.engine
+      if (!modelEngine) {
+        toaster({
+          title: 'Failed to start model',
+          description: `Engine for model ${model.model} is undefined`,
+          type: 'error',
+        })
+        return
+      }
+      if (!engineData) {
+        toaster({
+          title: 'Failed to start model',
+          description: `Engine data is not available. Please try again!`,
+          type: 'error',
+        })
+        return
+      }
+      const engineStatus = engineData.find((e) => e.name === modelEngine)
+      if (!engineStatus) {
+        toaster({
+          title: 'Failed to start model',
+          description: `Engine ${modelEngine} is not available`,
+          type: 'error',
+        })
+        console.error(`Engine ${modelEngine} is not available`)
+        return
+      }
+
+      if (
+        LocalEngines.find((e) => e === modelEngine) != null &&
+        engineStatus.status === 'not_initialized'
+      ) {
+        toaster({
+          title: 'Please wait for engine to initialize',
+          description: `Please retry after engine ${engineStatus.name} is installed.`,
+          type: 'default',
+        })
+        initializeEngine.mutate(modelEngine)
+        return
+      }
 
       if (activeModels.length >= concurrentModelWarningThreshold) {
         // if max concurrent models reached, stop the first model
@@ -88,6 +130,9 @@ const ModelItem: React.FC<Props> = ({ model }) => {
       stopModel,
       activeModels.length,
       setShowWarningMultipleModelModal,
+      engineData,
+      initializeEngine,
+      model,
     ]
   )
 
diff --git a/web/screens/Thread/ThreadCenterPanel/ChatBody/EmptyModel/OnDeviceListStarter.tsx b/web/screens/Thread/ThreadCenterPanel/ChatBody/EmptyModel/OnDeviceListStarter.tsx
index 997e57caec..fc202a0c01 100644
--- a/web/screens/Thread/ThreadCenterPanel/ChatBody/EmptyModel/OnDeviceListStarter.tsx
+++ b/web/screens/Thread/ThreadCenterPanel/ChatBody/EmptyModel/OnDeviceListStarter.tsx
@@ -53,11 +53,11 @@ const OnDeviceStarterScreen = () => {
     data.modelCategories.get('HuggingFace') || []
 
   const engineModelMap = new Map<typeof RemoteEngines, HfModelEntry[]>()
-  Object.entries(data.modelCategories).forEach(([key, value]) => {
+  for (const [key, value] of data.modelCategories) {
     if (key !== 'HuggingFace' && key !== 'BuiltInModels') {
       engineModelMap.set(key as unknown as typeof RemoteEngines, value)
     }
-  })
+  }
 
   const models: HfModelEntry[] = builtInModels.concat(huggingFaceModels)
 
diff --git a/web/screens/Thread/index.tsx b/web/screens/Thread/index.tsx
index a918b75e66..d3f41ac1d6 100644
--- a/web/screens/Thread/index.tsx
+++ b/web/screens/Thread/index.tsx
@@ -1,10 +1,14 @@
-import { Fragment, useEffect } from 'react'
+import { Fragment, useMemo } from 'react'
 
-import { Model } from '@janhq/core'
-import { useAtom, useAtomValue } from 'jotai'
+import {
+  EngineStatus,
+  LlmEngine,
+  RemoteEngine,
+  RemoteEngines,
+} from '@janhq/core'
+import { useAtomValue } from 'jotai'
 
-import useCortex from '@/hooks/useCortex'
-import useModels from '@/hooks/useModels'
+import useEngineQuery from '@/hooks/useEngineQuery'
 
 import ThreadLeftPanel from '@/screens/Thread/ThreadLeftPanel'
 
@@ -14,43 +18,44 @@ import ThreadRightPanel from './ThreadRightPanel'
 
 import { waitingForCortexAtom } from '@/helpers/atoms/App.atom'
 import { downloadedModelsAtom } from '@/helpers/atoms/Model.atom'
-import {
-  isAnyRemoteModelConfiguredAtom,
-  setUpRemoteModelStageAtom,
-} from '@/helpers/atoms/SetupRemoteModel.atom'
 
 const ThreadScreen = () => {
   const downloadedModels = useAtomValue(downloadedModelsAtom)
   const waitingForCortex = useAtomValue(waitingForCortexAtom)
-  const isAnyRemoteModelConfigured = useAtomValue(
-    isAnyRemoteModelConfiguredAtom
-  )
-  const { createModel } = useCortex()
-  const { getModels } = useModels()
-
-  const [{ metadata }] = useAtom(setUpRemoteModelStageAtom)
-
-  useEffect(() => {
-    if (isAnyRemoteModelConfigured) {
-      createModel(metadata?.model as Model)
-      getModels()
+  const { data: engineData } = useEngineQuery()
+
+  const isAnyRemoteModelConfigured = useMemo(() => {
+    if (!engineData) return false
+
+    let result = false
+    for (const engine of engineData) {
+      if (RemoteEngines.includes(engine.name as RemoteEngine)) {
+        if (engine.status === EngineStatus.Ready) {
+          result = true
+        }
+      }
     }
-    // eslint-disable-next-line react-hooks/exhaustive-deps
-  }, [isAnyRemoteModelConfigured])
+    return result
+  }, [engineData])
+
+  const shouldShowEmptyModel = useMemo(
+    () => !downloadedModels.length && !isAnyRemoteModelConfigured,
+    [downloadedModels, isAnyRemoteModelConfigured]
+  )
 
   if (waitingForCortex) return null
 
   return (
     <div className="relative flex h-full w-full flex-1 overflow-x-hidden">
-      {!downloadedModels.length && !isAnyRemoteModelConfigured ? (
+      {shouldShowEmptyModel ? (
         <EmptyModel />
       ) : (
         <Fragment>
           <ThreadLeftPanel />
           <ThreadCenterPanel />
+          <ThreadRightPanel />
         </Fragment>
       )}
-      <ThreadRightPanel />
     </div>
   )
 }
diff --git a/web/utils/huggingface.ts b/web/utils/huggingface.ts
index ef69195347..cae943b73d 100644
--- a/web/utils/huggingface.ts
+++ b/web/utils/huggingface.ts
@@ -106,7 +106,6 @@ export const fetchCortexHubModels = async (): Promise<HfModelEntry[]> => {
       }
     }
   }
-
   return modelEntries
 }
 
@@ -333,9 +332,8 @@ export const fetchHuggingFaceRepoData = async (
   return data
 }
 
-// TODO: move this to somewhere else
 export interface HfModelEntry extends ModelEntry {
-  model?: Model // TODO: deprecated this
+  model?: Model
   remoteModel?: boolean
   engine?: LlmEngine
 }