diff --git a/desktop/src/__tests__/pages.test.tsx b/desktop/src/__tests__/pages.test.tsx
index 4ea9864e..f26f7297 100644
--- a/desktop/src/__tests__/pages.test.tsx
+++ b/desktop/src/__tests__/pages.test.tsx
@@ -259,6 +259,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -302,6 +303,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -350,6 +352,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -417,6 +420,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -487,6 +491,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -543,6 +548,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -599,6 +605,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [
@@ -663,6 +670,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -748,6 +756,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 50_000, output_tokens: 1_000 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -804,6 +813,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -868,6 +878,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -948,6 +959,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1031,6 +1043,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 72_000, output_tokens: 2_000 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1100,6 +1113,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1197,6 +1211,7 @@ describe('Content-only pages render without errors', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 26_000, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/components/chat/AskUserQuestion.test.tsx b/desktop/src/components/chat/AskUserQuestion.test.tsx
index 9428c364..c6ecf7ad 100644
--- a/desktop/src/components/chat/AskUserQuestion.test.tsx
+++ b/desktop/src/components/chat/AskUserQuestion.test.tsx
@@ -63,6 +63,7 @@ describe('AskUserQuestion', () => {
           },
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/components/chat/ChatInput.test.tsx b/desktop/src/components/chat/ChatInput.test.tsx
index d95a66cb..95e34173 100644
--- a/desktop/src/components/chat/ChatInput.test.tsx
+++ b/desktop/src/components/chat/ChatInput.test.tsx
@@ -188,6 +188,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -255,6 +256,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -273,6 +275,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -430,6 +433,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -481,6 +485,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -527,6 +532,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -591,6 +597,7 @@ describe('ChatInput file mentions', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/components/chat/MessageList.test.tsx b/desktop/src/components/chat/MessageList.test.tsx
index e0edd062..928c838c 100644
--- a/desktop/src/components/chat/MessageList.test.tsx
+++ b/desktop/src/components/chat/MessageList.test.tsx
@@ -36,6 +36,7 @@ function makeSessionState(overrides: Partial<PerSessionState> = {}): PerSessionS
     pendingPermission: null,
     pendingComputerUsePermission: null,
     tokenUsage: { input_tokens: 0, output_tokens: 0 },
+    streamingResponseChars: 0,
     elapsedSeconds: 0,
     statusVerb: '',
     apiRetry: null,
@@ -580,7 +581,7 @@ describe('MessageList nested tool calls', () => {
     expect(card.textContent).toContain('Background command')
     expect(card.textContent).toContain('running')
     expect(card.textContent).toContain('Running Playwright checks')
-    expect(card.textContent).toContain('1,200 tokens')
+    expect(card.textContent).toContain('1.2k tokens')
     expect(card.textContent).toContain('45s')
   })
 
diff --git a/desktop/src/components/chat/MessageList.tsx b/desktop/src/components/chat/MessageList.tsx
index 72273f5f..a234f704 100644
--- a/desktop/src/components/chat/MessageList.tsx
+++ b/desktop/src/components/chat/MessageList.tsx
@@ -23,6 +23,7 @@ import { StreamingIndicator } from './StreamingIndicator'
 import { InlineTaskSummary } from './InlineTaskSummary'
 import { CurrentTurnChangeCard } from './CurrentTurnChangeCard'
 import type { AgentTaskNotification, UIMessage } from '../../types/chat'
+import { formatTokenCount } from '../../lib/formatTokenCount'
 import { ConfirmDialog } from '../shared/ConfirmDialog'
 import { clearWindowSelection, getSelectionPopoverPosition, useSelectionPopoverDismiss } from '../../hooks/useSelectionPopoverDismiss'
 import {
@@ -158,11 +159,6 @@ function ChatSelectionMenu({
   )
 }
 
-function formatCompactTokenCount(tokens: number): string {
-  if (tokens >= 1000) return `${Math.round(tokens / 100) / 10}k`
-  return String(tokens)
-}
-
 function getCompactSummaryTitle(message: CompactSummaryEvent, t: ReturnType<typeof useTranslation>) {
   if (message.trigger === 'auto') return t('chat.compactSummary.autoTitle')
   if (message.trigger === 'manual') return t('chat.compactSummary.manualTitle')
@@ -179,7 +175,7 @@ function CompactStatusDivider({ message, state }: { message?: CompactSummaryEven
   const meta = [
     message?.trigger ? t(`chat.compactSummary.trigger.${message.trigger}` as TranslationKey) : null,
     typeof message?.preTokens === 'number'
-      ? t('chat.compactSummary.tokens', { count: formatCompactTokenCount(message.preTokens) })
+      ? t('chat.compactSummary.tokens', { count: formatTokenCount(message.preTokens) })
       : null,
     typeof message?.messagesSummarized === 'number'
       ? t('chat.compactSummary.messages', { count: String(message.messagesSummarized) })
@@ -350,7 +346,7 @@ function BackgroundTaskEventCard({ message }: { message: BackgroundTaskEvent })
             </span>
             {task.usage?.totalTokens ? (
               <span className="hidden shrink-0 text-[11px] text-[var(--color-text-tertiary)] sm:inline">
-                {t('chat.backgroundAgents.tokens', { count: task.usage.totalTokens.toLocaleString() })}
+                {t('chat.backgroundAgents.tokens', { count: formatTokenCount(task.usage.totalTokens) })}
               </span>
             ) : null}
             {duration ? (
diff --git a/desktop/src/components/chat/PlanModePermissionDialog.test.tsx b/desktop/src/components/chat/PlanModePermissionDialog.test.tsx
index c76908fc..458639fd 100644
--- a/desktop/src/components/chat/PlanModePermissionDialog.test.tsx
+++ b/desktop/src/components/chat/PlanModePermissionDialog.test.tsx
@@ -57,6 +57,7 @@ function seedPendingPlanPermission() {
         },
         pendingComputerUsePermission: null,
         tokenUsage: { input_tokens: 0, output_tokens: 0 },
+        streamingResponseChars: 0,
         elapsedSeconds: 0,
         statusVerb: '',
         slashCommands: [],
diff --git a/desktop/src/components/chat/StreamingIndicator.test.tsx b/desktop/src/components/chat/StreamingIndicator.test.tsx
new file mode 100644
index 00000000..ba32c732
--- /dev/null
+++ b/desktop/src/components/chat/StreamingIndicator.test.tsx
@@ -0,0 +1,94 @@
+import { beforeEach, describe, expect, it, vi } from 'vitest'
+import { render, screen } from '@testing-library/react'
+
+vi.mock('../../api/websocket', () => ({
+  wsManager: {
+    connect: vi.fn(),
+    disconnect: vi.fn(),
+    onMessage: vi.fn(() => () => {}),
+    clearHandlers: vi.fn(),
+    send: vi.fn(),
+  },
+}))
+
+vi.mock('../../api/sessions', () => ({
+  sessionsApi: {
+    getMessages: vi.fn(async () => ({ messages: [] })),
+    getSlashCommands: vi.fn(async () => ({ commands: [] })),
+  },
+}))
+
+import { StreamingIndicator } from './StreamingIndicator'
+import { useChatStore, type PerSessionState } from '../../stores/chatStore'
+import { useSettingsStore } from '../../stores/settingsStore'
+import { useTabStore } from '../../stores/tabStore'
+
+const ACTIVE_TAB = 'active-tab'
+
+function makeSession(overrides: Partial<PerSessionState> = {}): PerSessionState {
+  return {
+    messages: [],
+    chatState: 'streaming',
+    connectionState: 'connected',
+    historyStatus: 'idle',
+    historyError: null,
+    streamingText: '',
+    streamingToolInput: '',
+    activeToolUseId: null,
+    activeToolName: null,
+    activeThinkingId: null,
+    pendingPermission: null,
+    pendingComputerUsePermission: null,
+    tokenUsage: { input_tokens: 0, output_tokens: 0 },
+    streamingResponseChars: 0,
+    elapsedSeconds: 0,
+    statusVerb: '',
+    apiRetry: null,
+    slashCommands: [],
+    agentTaskNotifications: {},
+    backgroundAgentTasks: {},
+    elapsedTimer: null,
+    ...overrides,
+  }
+}
+
+describe('StreamingIndicator', () => {
+  beforeEach(() => {
+    useSettingsStore.setState({ locale: 'en' })
+    useTabStore.setState({
+      activeTabId: ACTIVE_TAB,
+      tabs: [{ sessionId: ACTIVE_TAB, title: 'Test', type: 'session', status: 'running' }],
+    })
+  })
+
+  // issue #757: token usage is rendered with the shared compact notation and
+  // an explicit unit instead of a bare number.
+  it('renders the current turn token estimate as "↓ N tokens"', () => {
+    useChatStore.setState({
+      sessions: {
+        // 8976 streamed chars ÷ 4 = 2244 tokens → "2.2k tokens"
+        [ACTIVE_TAB]: makeSession({ streamingResponseChars: 8976 }),
+      },
+    })
+
+    render(<StreamingIndicator />)
+
+    expect(screen.getByText(/↓ 2\.2k tokens/)).toBeTruthy()
+  })
+
+  it('hides the token estimate until this turn has streamed output', () => {
+    useChatStore.setState({
+      sessions: {
+        // Previous-turn usage must not leak into the indicator (issue #757
+        // follow-up report: the count showed a stale value from the last turn).
+        [ACTIVE_TAB]: makeSession({
+          tokenUsage: { input_tokens: 5000, output_tokens: 6240 },
+        }),
+      },
+    })
+
+    render(<StreamingIndicator />)
+
+    expect(screen.queryByText(/tokens/)).toBeNull()
+  })
+})
diff --git a/desktop/src/components/chat/StreamingIndicator.tsx b/desktop/src/components/chat/StreamingIndicator.tsx
index a2e35d70..6d885647 100644
--- a/desktop/src/components/chat/StreamingIndicator.tsx
+++ b/desktop/src/components/chat/StreamingIndicator.tsx
@@ -3,6 +3,7 @@ import { RefreshCw } from 'lucide-react'
 import { useChatStore } from '../../stores/chatStore'
 import { useTabStore } from '../../stores/tabStore'
 import { useTranslation, type TranslationKey } from '../../i18n'
+import { formatTokenCount } from '../../lib/formatTokenCount'
 
 function formatElapsed(seconds: number): string {
   if (seconds < 60) return `${seconds}s`
@@ -41,7 +42,8 @@ export function StreamingIndicator() {
   const statusVerb = sessionState?.statusVerb ?? ''
   const apiRetry = sessionState?.apiRetry ?? null
   const elapsedSeconds = sessionState?.elapsedSeconds ?? 0
-  const tokenUsage = sessionState?.tokenUsage ?? { input_tokens: 0, output_tokens: 0 }
+  // chars ÷ 4 estimates output tokens for this turn, mirroring the CLI spinner.
+  const streamingTokens = Math.round((sessionState?.streamingResponseChars ?? 0) / 4)
 
   useEffect(() => {
     if (!apiRetry) return undefined
@@ -106,9 +108,9 @@ export function StreamingIndicator() {
           {formatElapsed(elapsedSeconds)}
         </span>
       )}
-      {tokenUsage.output_tokens > 0 && (
+      {streamingTokens > 0 && (
         <span className="text-[10px] text-[var(--color-text-tertiary)]">
-          · ↓ {tokenUsage.output_tokens}
+          · ↓ {t('common.tokens', { count: formatTokenCount(streamingTokens) })}
         </span>
       )}
     </div>
diff --git a/desktop/src/components/chat/chatBlocks.test.tsx b/desktop/src/components/chat/chatBlocks.test.tsx
index ad7d05be..9f825c21 100644
--- a/desktop/src/components/chat/chatBlocks.test.tsx
+++ b/desktop/src/components/chat/chatBlocks.test.tsx
@@ -281,6 +281,7 @@ describe('chat blocks', () => {
           },
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/components/layout/TabBar.test.tsx b/desktop/src/components/layout/TabBar.test.tsx
index 0fee008e..c9ac1e20 100644
--- a/desktop/src/components/layout/TabBar.test.tsx
+++ b/desktop/src/components/layout/TabBar.test.tsx
@@ -30,6 +30,7 @@ function makeChatSession(chatState: ChatState): PerSessionState {
     pendingPermission: null,
     pendingComputerUsePermission: null,
     tokenUsage: { input_tokens: 0, output_tokens: 0 },
+    streamingResponseChars: 0,
     elapsedSeconds: 0,
     statusVerb: '',
     slashCommands: [],
diff --git a/desktop/src/components/workspace/WorkspacePanel.test.tsx b/desktop/src/components/workspace/WorkspacePanel.test.tsx
index 6d3a64d9..71f30172 100644
--- a/desktop/src/components/workspace/WorkspacePanel.test.tsx
+++ b/desktop/src/components/workspace/WorkspacePanel.test.tsx
@@ -1369,6 +1369,7 @@ describe('WorkspacePanel', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/i18n/locales/en.ts b/desktop/src/i18n/locales/en.ts
index 8722011a..d26d31d8 100644
--- a/desktop/src/i18n/locales/en.ts
+++ b/desktop/src/i18n/locales/en.ts
@@ -7,6 +7,7 @@ export const en = {
   'common.run': 'Run',
   'common.send': 'Send',
   'common.stop': 'Stop',
+  'common.tokens': '{count} tokens',
   'common.rename': 'Rename',
   'common.retry': 'Retry',
   'common.loading': 'Loading...',
diff --git a/desktop/src/i18n/locales/jp.ts b/desktop/src/i18n/locales/jp.ts
index 00f696af..19a4e515 100644
--- a/desktop/src/i18n/locales/jp.ts
+++ b/desktop/src/i18n/locales/jp.ts
@@ -9,6 +9,7 @@ export const jp: Record<TranslationKey, string> = {
   'common.run': '実行',
   'common.send': '送信',
   'common.stop': '停止',
+  'common.tokens': '{count} トークン',
   'common.rename': '名前を変更',
   'common.retry': '再試行',
   'common.loading': '読み込み中...',
diff --git a/desktop/src/i18n/locales/kr.ts b/desktop/src/i18n/locales/kr.ts
index c401d458..47e52ff2 100644
--- a/desktop/src/i18n/locales/kr.ts
+++ b/desktop/src/i18n/locales/kr.ts
@@ -9,6 +9,7 @@ export const kr: Record<TranslationKey, string> = {
   'common.run': '실행',
   'common.send': '전송',
   'common.stop': '중지',
+  'common.tokens': '{count} 토큰',
   'common.rename': '이름 바꾸기',
   'common.retry': '다시 시도',
   'common.loading': '불러오는 중...',
diff --git a/desktop/src/i18n/locales/zh-TW.ts b/desktop/src/i18n/locales/zh-TW.ts
index 7a218ddd..fafbf063 100644
--- a/desktop/src/i18n/locales/zh-TW.ts
+++ b/desktop/src/i18n/locales/zh-TW.ts
@@ -9,6 +9,7 @@ export const zh: Record<TranslationKey, string> = {
   'common.run': '執行',
   'common.send': '傳送',
   'common.stop': '停止',
+  'common.tokens': '{count} tokens',
   'common.rename': '重新命名',
   'common.retry': '重試',
   'common.loading': '載入中...',
diff --git a/desktop/src/i18n/locales/zh.ts b/desktop/src/i18n/locales/zh.ts
index d0e6822f..fa1062b1 100644
--- a/desktop/src/i18n/locales/zh.ts
+++ b/desktop/src/i18n/locales/zh.ts
@@ -9,6 +9,7 @@ export const zh: Record<TranslationKey, string> = {
   'common.run': '运行',
   'common.send': '发送',
   'common.stop': '停止',
+  'common.tokens': '{count} tokens',
   'common.rename': '重命名',
   'common.retry': '重试',
   'common.loading': '加载中...',
diff --git a/desktop/src/lib/formatTokenCount.test.ts b/desktop/src/lib/formatTokenCount.test.ts
new file mode 100644
index 00000000..5edcf708
--- /dev/null
+++ b/desktop/src/lib/formatTokenCount.test.ts
@@ -0,0 +1,30 @@
+import { describe, expect, it } from 'vitest'
+import { formatTokenCount } from './formatTokenCount'
+
+describe('formatTokenCount', () => {
+  it('formats missing or invalid values as --', () => {
+    expect(formatTokenCount()).toBe('--')
+    expect(formatTokenCount(Number.NaN)).toBe('--')
+    expect(formatTokenCount(-1)).toBe('--')
+  })
+
+  it('formats counts below 1000 verbatim', () => {
+    expect(formatTokenCount(0)).toBe('0')
+    expect(formatTokenCount(847)).toBe('847')
+  })
+
+  it('formats thousands with one decimal and a k suffix', () => {
+    expect(formatTokenCount(1234)).toBe('1.2k')
+    expect(formatTokenCount(38_500)).toBe('38.5k')
+    expect(formatTokenCount(124_320)).toBe('124.3k')
+  })
+
+  it('drops the trailing .0 to match the CLI notation', () => {
+    expect(formatTokenCount(1000)).toBe('1k')
+    expect(formatTokenCount(2_000_000)).toBe('2m')
+  })
+
+  it('formats millions with an m suffix', () => {
+    expect(formatTokenCount(2_400_000)).toBe('2.4m')
+  })
+})
diff --git a/desktop/src/lib/formatTokenCount.ts b/desktop/src/lib/formatTokenCount.ts
new file mode 100644
index 00000000..496b2fce
--- /dev/null
+++ b/desktop/src/lib/formatTokenCount.ts
@@ -0,0 +1,13 @@
+/**
+ * Compact token-count formatter shared by the chat UI (session header,
+ * streaming indicator, compact summary, background agents) and trace views,
+ * so every surface renders token usage with one notation.
+ * "847" below 1000, "1.2k" up to 1M, "1.2m" beyond — trailing ".0" dropped
+ * ("1k", not "1.0k") to match the CLI's formatTokens.
+ */
+export function formatTokenCount(n?: number): string {
+  if (typeof n !== 'number' || !Number.isFinite(n) || n < 0) return '--'
+  if (n < 1000) return String(Math.round(n))
+  if (n < 1_000_000) return `${(n / 1000).toFixed(1).replace(/\.0$/, '')}k`
+  return `${(n / 1_000_000).toFixed(1).replace(/\.0$/, '')}m`
+}
diff --git a/desktop/src/lib/trace/formatters.test.ts b/desktop/src/lib/trace/formatters.test.ts
index f6288751..c37876ed 100644
--- a/desktop/src/lib/trace/formatters.test.ts
+++ b/desktop/src/lib/trace/formatters.test.ts
@@ -29,25 +29,10 @@ describe('formatDurationMs', () => {
 })
 
 describe('formatTokenCount', () => {
-  it('formats missing or invalid values as --', () => {
-    expect(formatTokenCount()).toBe('--')
-    expect(formatTokenCount(Number.NaN)).toBe('--')
-    expect(formatTokenCount(-1)).toBe('--')
-  })
-
-  it('formats counts below 1000 verbatim', () => {
-    expect(formatTokenCount(0)).toBe('0')
-    expect(formatTokenCount(847)).toBe('847')
-  })
-
-  it('formats thousands with one decimal and a k suffix', () => {
-    expect(formatTokenCount(1000)).toBe('1.0k')
+  // Full coverage lives in ../formatTokenCount.test.ts; this verifies the re-export.
+  it('re-exports the shared compact formatter', () => {
     expect(formatTokenCount(1234)).toBe('1.2k')
-    expect(formatTokenCount(38_500)).toBe('38.5k')
-  })
-
-  it('formats millions with an m suffix', () => {
-    expect(formatTokenCount(2_400_000)).toBe('2.4m')
+    expect(formatTokenCount(1000)).toBe('1k')
   })
 })
 
diff --git a/desktop/src/lib/trace/formatters.ts b/desktop/src/lib/trace/formatters.ts
index 93cb19c2..db9db257 100644
--- a/desktop/src/lib/trace/formatters.ts
+++ b/desktop/src/lib/trace/formatters.ts
@@ -1,6 +1,9 @@
 import type { TraceCallUsage } from '../../types/trace'
+import { formatTokenCount } from '../formatTokenCount'
 import type { NormalizedUsage } from './types'
 
+export { formatTokenCount }
+
 export function formatDurationMs(ms?: number): string {
   if (typeof ms !== 'number' || !Number.isFinite(ms) || ms < 0) return '--'
   if (ms < 1000) return `${Math.round(ms)}ms`
@@ -12,13 +15,6 @@ export function formatDurationMs(ms?: number): string {
   return `${minutes}m ${String(seconds).padStart(2, '0')}s`
 }
 
-export function formatTokenCount(n?: number): string {
-  if (typeof n !== 'number' || !Number.isFinite(n) || n < 0) return '--'
-  if (n < 1000) return String(Math.round(n))
-  if (n < 1_000_000) return `${(n / 1000).toFixed(1)}k`
-  return `${(n / 1_000_000).toFixed(1)}m`
-}
-
 export function formatUsageBrief(u?: NormalizedUsage | TraceCallUsage): string {
   if (!u) return '--'
   return `${formatTokenCount(u.inputTokens)} → ${formatTokenCount(u.outputTokens)}`
diff --git a/desktop/src/pages/ActiveSession.test.tsx b/desktop/src/pages/ActiveSession.test.tsx
index e7f63bdd..c1ac306d 100644
--- a/desktop/src/pages/ActiveSession.test.tsx
+++ b/desktop/src/pages/ActiveSession.test.tsx
@@ -132,6 +132,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -239,6 +240,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -305,6 +307,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -379,6 +382,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -441,6 +445,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -502,6 +507,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -581,6 +587,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -636,6 +643,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -708,6 +716,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -763,6 +772,7 @@ describe('ActiveSession task polling', () => {
             pendingPermission: null,
             pendingComputerUsePermission: null,
             tokenUsage: { input_tokens: 0, output_tokens: 0 },
+            streamingResponseChars: 0,
             elapsedSeconds: 0,
             statusVerb: '',
             slashCommands: [],
@@ -816,6 +826,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -875,6 +886,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -982,6 +994,7 @@ describe('ActiveSession task polling', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
diff --git a/desktop/src/pages/ActiveSession.tsx b/desktop/src/pages/ActiveSession.tsx
index b49a19f8..39cd5cc5 100644
--- a/desktop/src/pages/ActiveSession.tsx
+++ b/desktop/src/pages/ActiveSession.tsx
@@ -31,6 +31,7 @@ import type { SessionListItem } from '../types/session'
 import type { ActiveGoalState } from '../types/chat'
 import { useMobileViewport } from '../hooks/useMobileViewport'
 import { isDesktopRuntime } from '../lib/desktopRuntime'
+import { formatTokenCount } from '../lib/formatTokenCount'
 import { publicAssetPath } from '../lib/publicAsset'
 
 const TASK_POLL_INTERVAL_MS = 1000
@@ -479,7 +480,9 @@ export function ActiveSession() {
                       {totalTokens > 0 && (
                         <>
                           <span className="text-[var(--color-outline)]">·</span>
-                          <span>{totalTokens.toLocaleString()} t</span>
+                          <span title={t('common.tokens', { count: totalTokens.toLocaleString() })}>
+                            {t('common.tokens', { count: formatTokenCount(totalTokens) })}
+                          </span>
                         </>
                       )}
                       {lastUpdated && (
diff --git a/desktop/src/stores/chatStore.test.ts b/desktop/src/stores/chatStore.test.ts
index ced3ef43..53998b03 100644
--- a/desktop/src/stores/chatStore.test.ts
+++ b/desktop/src/stores/chatStore.test.ts
@@ -151,6 +151,7 @@ function makeSession(overrides: Partial<PerSessionState> = {}): PerSessionState
     pendingPermission: null,
     pendingComputerUsePermission: null,
     tokenUsage: { input_tokens: 0, output_tokens: 0 },
+    streamingResponseChars: 0,
     elapsedSeconds: 0,
     statusVerb: '',
     apiRetry: null,
@@ -1292,6 +1293,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1376,6 +1378,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1452,6 +1455,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -1619,6 +1623,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [{ name: 'old-command', description: 'Old command' }],
@@ -1992,6 +1997,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2055,6 +2061,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2116,6 +2123,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2437,6 +2445,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 12, output_tokens: 34 },
+          streamingResponseChars: 999,
           elapsedSeconds: 5,
           statusVerb: 'Thinking',
           slashCommands: [],
@@ -2461,6 +2470,7 @@ describe('chatStore history mapping', () => {
     expect(session?.streamingText).toBe('')
     expect(session?.chatState).toBe('idle')
     expect(session?.tokenUsage).toEqual({ input_tokens: 0, output_tokens: 0 })
+    expect(session?.streamingResponseChars).toBe(0)
     expect(session?.slashCommands).toEqual([])
     expect(clearTasksMock).toHaveBeenCalledWith(TEST_SESSION_ID)
 
@@ -2510,6 +2520,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2555,6 +2566,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: 'Compacting conversation',
           slashCommands: [],
@@ -2609,6 +2621,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2657,6 +2670,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2737,6 +2751,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -2981,6 +2996,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3026,6 +3042,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3133,6 +3150,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3203,6 +3221,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3493,6 +3512,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3595,6 +3615,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3647,6 +3668,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3679,6 +3701,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3720,6 +3743,7 @@ describe('chatStore history mapping', () => {
             },
           },
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3784,6 +3808,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3825,6 +3850,7 @@ describe('chatStore history mapping', () => {
           pendingPermission: null,
           pendingComputerUsePermission: null,
           tokenUsage: { input_tokens: 0, output_tokens: 0 },
+          streamingResponseChars: 0,
           elapsedSeconds: 0,
           statusVerb: '',
           slashCommands: [],
@@ -3861,4 +3887,57 @@ describe('chatStore history mapping', () => {
       attachments: undefined,
     })
   })
+
+  // issue #757: the streaming indicator estimates this turn's output tokens
+  // from streamed characters (÷4, mirroring the CLI spinner) instead of
+  // showing the previous turn's stale usage.
+  it('accumulates streamed text, tool input, and thinking chars for the token estimate', () => {
+    vi.useFakeTimers()
+    useChatStore.setState({ sessions: { [TEST_SESSION_ID]: makeSession() } })
+
+    const charsOf = () =>
+      useChatStore.getState().sessions[TEST_SESSION_ID]?.streamingResponseChars
+
+    useChatStore.getState().handleServerMessage(TEST_SESSION_ID, {
+      type: 'content_delta',
+      text: 'a'.repeat(40),
+    })
+    vi.advanceTimersByTime(60)
+    expect(charsOf()).toBe(40)
+
+    useChatStore.getState().handleServerMessage(TEST_SESSION_ID, {
+      type: 'content_delta',
+      toolInput: '{"a":1}',
+    })
+    vi.advanceTimersByTime(60)
+    expect(charsOf()).toBe(47)
+
+    useChatStore.getState().handleServerMessage(TEST_SESSION_ID, {
+      type: 'thinking',
+      text: 'pondering',
+    })
+    expect(charsOf()).toBe(56)
+
+    vi.runOnlyPendingTimers()
+    vi.useRealTimers()
+  })
+
+  it('resets the streaming token estimate when the user sends the next message', () => {
+    vi.useFakeTimers()
+    useChatStore.setState({
+      sessions: {
+        [TEST_SESSION_ID]: makeSession({
+          chatState: 'idle',
+          streamingResponseChars: 4321,
+        }),
+      },
+    })
+
+    useChatStore.getState().sendMessage(TEST_SESSION_ID, '继续')
+
+    const session = useChatStore.getState().sessions[TEST_SESSION_ID]
+    expect(session?.streamingResponseChars).toBe(0)
+    if (session?.elapsedTimer) clearInterval(session.elapsedTimer)
+    vi.useRealTimers()
+  })
 })
diff --git a/desktop/src/stores/chatStore.ts b/desktop/src/stores/chatStore.ts
index fc1dd402..a9cdb4ea 100644
--- a/desktop/src/stores/chatStore.ts
+++ b/desktop/src/stores/chatStore.ts
@@ -79,6 +79,12 @@ export type PerSessionState = {
     request: ComputerUsePermissionRequest
   } | null
   tokenUsage: TokenUsage
+  /**
+   * Characters streamed by the assistant during the current turn (text,
+   * thinking, tool input). ÷4 approximates output tokens for the streaming
+   * indicator — same estimation the CLI spinner uses. Reset on each send.
+   */
+  streamingResponseChars: number
   elapsedSeconds: number
   statusVerb: string
   apiRetry?: ApiRetryState | null
@@ -111,6 +117,7 @@ const DEFAULT_SESSION_STATE: PerSessionState = {
   pendingPermission: null,
   pendingComputerUsePermission: null,
   tokenUsage: { input_tokens: 0, output_tokens: 0 },
+  streamingResponseChars: 0,
   elapsedSeconds: 0,
   statusVerb: '',
   apiRetry: null,
@@ -966,6 +973,7 @@ export const useChatStore = create<ChatStore>((set, get) => ({
             chatState: 'thinking',
             elapsedSeconds: 0,
             streamingText: '',
+            streamingResponseChars: 0,
             statusVerb: isMemberSession ? '' : randomSpinnerVerb(),
             apiRetry: null,
             elapsedTimer: timer,
@@ -1327,7 +1335,6 @@ export const useChatStore = create<ChatStore>((set, get) => ({
               : msg.verb && msg.verb !== 'Thinking'
                 ? msg.verb
                 : '',
-            ...(msg.tokens ? { tokenUsage: { ...session.tokenUsage, output_tokens: msg.tokens } } : {}),
             ...(msg.state === 'idle' ? { activeThinkingId: null } : {}),
             ...(msg.state === 'idle' ? { apiRetry: null } : {}),
             ...(nextMessages !== session.messages ? { messages: nextMessages } : {}),
@@ -1439,7 +1446,10 @@ export const useChatStore = create<ChatStore>((set, get) => ({
               const text = pendingDeltaBySession.get(sessionId) ?? ''
               pendingDeltaBySession.delete(sessionId)
               flushTimerBySession.delete(sessionId)
-              update((s) => ({ streamingText: s.streamingText + text }))
+              update((s) => ({
+                streamingText: s.streamingText + text,
+                streamingResponseChars: s.streamingResponseChars + text.length,
+              }))
             }, 50)
             flushTimerBySession.set(sessionId, timer)
           }
@@ -1455,6 +1465,7 @@ export const useChatStore = create<ChatStore>((set, get) => ({
                 const activeToolUseId = s.activeToolUseId
                 return {
                   streamingToolInput: partialInput,
+                  streamingResponseChars: s.streamingResponseChars + text.length,
                   ...(activeToolUseId
                     ? {
                         messages: upsertToolUseMessage(s.messages, activeToolUseId, (existing) => {
@@ -1491,7 +1502,13 @@ export const useChatStore = create<ChatStore>((set, get) => ({
           if (last && last.type === 'thinking') {
             const updated = [...base]
             updated[updated.length - 1] = { ...last, content: last.content + msg.text }
-            return { messages: updated, chatState: 'thinking', activeThinkingId: last.id, streamingText: '' }
+            return {
+              messages: updated,
+              chatState: 'thinking',
+              activeThinkingId: last.id,
+              streamingText: '',
+              streamingResponseChars: s.streamingResponseChars + msg.text.length,
+            }
           }
           const id = nextId()
           return {
@@ -1499,6 +1516,7 @@ export const useChatStore = create<ChatStore>((set, get) => ({
             chatState: 'thinking',
             activeThinkingId: id,
             streamingText: '',
+            streamingResponseChars: s.streamingResponseChars + msg.text.length,
           }
         })
         break
@@ -1767,6 +1785,7 @@ export const useChatStore = create<ChatStore>((set, get) => ({
             statusVerb: '',
             apiRetry: null,
             tokenUsage: { input_tokens: 0, output_tokens: 0 },
+            streamingResponseChars: 0,
             slashCommands: [],
             activeGoal: null,
             backgroundAgentTasks: {},
diff --git a/desktop/src/stores/teamStore.ts b/desktop/src/stores/teamStore.ts
index ce6d7e71..aad9cfbf 100644
--- a/desktop/src/stores/teamStore.ts
+++ b/desktop/src/stores/teamStore.ts
@@ -29,6 +29,7 @@ function createMemberSessionState() {
     pendingPermission: null,
     pendingComputerUsePermission: null,
     tokenUsage: { input_tokens: 0, output_tokens: 0 },
+    streamingResponseChars: 0,
     elapsedSeconds: 0,
     statusVerb: '',
     slashCommands: [],
diff --git a/desktop/src/types/chat.ts b/desktop/src/types/chat.ts
index 8ee3a3f2..80609f40 100644
--- a/desktop/src/types/chat.ts
+++ b/desktop/src/types/chat.ts
@@ -94,7 +94,7 @@ export type ServerMessage =
     }
   | { type: 'message_complete'; usage: TokenUsage }
   | { type: 'thinking'; text: string }
-  | { type: 'status'; state: ChatState; verb?: string; elapsed?: number; tokens?: number }
+  | { type: 'status'; state: ChatState; verb?: string }
   // CLI 回传的权限模式变化（如 ExitPlanMode 退出 plan 后恢复、Shift+Tab）。
   // 桌面端据此把选择器校正回 CLI 的真实权限，避免本地影子值漂移。
   | { type: 'permission_mode_changed'; mode: PermissionMode }
diff --git a/src/components/Spinner/SpinnerAnimationRow.tsx b/src/components/Spinner/SpinnerAnimationRow.tsx
index 4e77bf90..74bff9e2 100644
--- a/src/components/Spinner/SpinnerAnimationRow.tsx
+++ b/src/components/Spinner/SpinnerAnimationRow.tsx
@@ -5,7 +5,7 @@ import { useMemo, useRef } from 'react';
 import { stringWidth } from '../../ink/stringWidth.js';
 import { Box, Text, useAnimationFrame } from '../../ink.js';
 import type { InProcessTeammateTaskState } from '../../tasks/InProcessTeammateTask/types.js';
-import { formatDuration, formatNumber } from '../../utils/format.js';
+import { formatDuration, formatTokens } from '../../utils/format.js';
 import { toInkColor } from '../../utils/ink.js';
 import type { Theme } from '../../utils/theme.js';
 import { Byline } from '../design-system/Byline.js';
@@ -164,7 +164,7 @@ export function SpinnerAnimationRow({
 
   // === Token count (leader + teammates, or foregrounded teammate) ===
   const totalTokens = foregroundedTeammate && !foregroundedTeammate.isIdle ? foregroundedTeammate.progress?.tokenCount ?? 0 : leaderTokens + teammateTokens;
-  const tokenCount = formatNumber(totalTokens);
+  const tokenCount = formatTokens(totalTokens);
   const tokensText = hasRunningTeammates ? `${tokenCount} tokens` : `${figures.arrowDown} ${tokenCount} tokens`;
   const tokensWidth = stringWidth(tokensText);
 
diff --git a/src/components/Spinner/TeammateSpinnerLine.tsx b/src/components/Spinner/TeammateSpinnerLine.tsx
index 638667be..e94b83d2 100644
--- a/src/components/Spinner/TeammateSpinnerLine.tsx
+++ b/src/components/Spinner/TeammateSpinnerLine.tsx
@@ -10,7 +10,7 @@ import { stringWidth } from '../../ink/stringWidth.js';
 import { Box, Text } from '../../ink.js';
 import type { InProcessTeammateTaskState } from '../../tasks/InProcessTeammateTask/types.js';
 import { summarizeRecentActivities } from '../../utils/collapseReadSearch.js';
-import { formatDuration, formatNumber, truncateToWidth } from '../../utils/format.js';
+import { formatDuration, formatTokens, truncateToWidth } from '../../utils/format.js';
 import { toInkColor } from '../../utils/ink.js';
 import { TEAMMATE_SELECT_HINT } from './teammateSelectHint.js';
 type Props = {
@@ -127,7 +127,7 @@ export function TeammateSpinnerLine({
   // Get stats from progress
   const toolUseCount = teammate.progress?.toolUseCount ?? 0;
   const tokenCount = teammate.progress?.tokenCount ?? 0;
-  const statsText = ` · ${toolUseCount} tool ${toolUseCount === 1 ? 'use' : 'uses'} · ${formatNumber(tokenCount)} tokens`;
+  const statsText = ` · ${toolUseCount} tool ${toolUseCount === 1 ? 'use' : 'uses'} · ${formatTokens(tokenCount)} tokens`;
   const statsWidth = stringWidth(statsText);
   const selectHintText = ` · ${TEAMMATE_SELECT_HINT}`;
   const selectHintWidth = stringWidth(selectHintText);
@@ -216,7 +216,7 @@ export function TeammateSpinnerLine({
         {showStats && <Text dimColor>
             {' '}
             · {toolUseCount} tool {toolUseCount === 1 ? 'use' : 'uses'} ·{' '}
-            {formatNumber(tokenCount)} tokens
+            {formatTokens(tokenCount)} tokens
           </Text>}
         {/* Hints: select hint when highlighted, view hint when selected but not foregrounded */}
         {showSelectHint && <Text dimColor> · {TEAMMATE_SELECT_HINT}</Text>}
diff --git a/src/server/ws/events.ts b/src/server/ws/events.ts
index e27dfe63..7f0e88d7 100644
--- a/src/server/ws/events.ts
+++ b/src/server/ws/events.ts
@@ -64,7 +64,7 @@ export type ServerMessage =
     }
   | { type: 'message_complete'; usage: TokenUsage }
   | { type: 'thinking'; text: string }
-  | { type: 'status'; state: ChatState; verb?: string; elapsed?: number; tokens?: number }
+  | { type: 'status'; state: ChatState; verb?: string }
   // CLI 是权限模式的唯一真相来源。当 CLI 内部 mode 变化（如 ExitPlanMode 后
   // 恢复到进入 plan 前的模式、Shift+Tab 切换）时，把新模式回传给前端，让桌面端
   // 选择器与 CLI 保持同步，而不是停留在本地影子值上。
diff --git a/tests/formatTokens.test.ts b/tests/formatTokens.test.ts
new file mode 100644
index 00000000..b4783991
--- /dev/null
+++ b/tests/formatTokens.test.ts
@@ -0,0 +1,23 @@
+import { describe, expect, test } from "bun:test"
+import { formatNumber, formatTokens } from "../src/utils/format.js"
+
+// The spinner rows render token counts via formatTokens (issue #757): same
+// compact notation as formatNumber, but without the ".0" artifact ("1k", not
+// "1.0k").
+describe("formatTokens", () => {
+  test("keeps counts below 1000 verbatim", () => {
+    expect(formatTokens(0)).toBe("0")
+    expect(formatTokens(999)).toBe("999")
+  })
+
+  test("compacts thousands with one decimal", () => {
+    expect(formatTokens(1234)).toBe("1.2k")
+    expect(formatTokens(38_500)).toBe("38.5k")
+  })
+
+  test("drops the trailing .0 that formatNumber keeps", () => {
+    expect(formatNumber(1000)).toBe("1.0k")
+    expect(formatTokens(1000)).toBe("1k")
+    expect(formatTokens(2_000_000)).toBe("2m")
+  })
+})