@@ -2030,7 +2030,6 @@ function AssistantFooter(props: { message: SessionMessageAssistant }) {
   const turn = () => ({ index: ctx.messageIndex(props.message.id), legacy: ctx.legacyTurns() })
   const duration = createMemo(() => SessionTurn.duration(messages(), props.message, turn()))
   const tokensPerSecond = createMemo(() => SessionTurn.tokensPerSecond(messages(), props.message, turn()))
-  const timeToFirstToken = createMemo(() => SessionTurn.timeToFirstToken(messages(), props.message, turn()))
   const interrupted = createMemo(() => props.message.error?.message === "Step interrupted")
   const freeLimit = createMemo(() => props.message.error?.type === "provider.free-limit")
   const limitNotice = useLimitNotice(() => props.message)
@@ -2065,9 +2064,6 @@ function AssistantFooter(props: { message: SessionMessageAssistant }) {
           <Show when={config.data.session.tps && tokensPerSecond()}>
             {(value) => <span style={{ fg: theme.text.muted }}> · {value().toFixed(1)} tok/s</span>}
           </Show>
-          <Show when={config.data.session.tps && ctx.terminal.width >= 36 && timeToFirstToken()}>
-            {(value) => <span style={{ fg: theme.text.muted }}> · {Locale.duration(value())} ttft</span>}
-          </Show>
           <Show when={interrupted()}>
             <span style={{ fg: theme.text.muted }}> · interrupted</span>
           </Show>
 
@@ -795,7 +795,8 @@ test("keeps assistant footer metrics current after prepend, same-length refresh,
     (frame) => frame.includes("Original answer") && frame.includes("40.0 tok/s"),
   )
   expect(initial).toContain("Current input")
-  expect(initial).toContain("4.0s \u00b7 40.0 tok/s \u00b7 500ms ttft")
+  expect(initial).toContain("4.0s \u00b7 40.0 tok/s")
+  expect(initial).not.toContain("ttft")
   expect(initial).not.toContain("Prepended input")
 
   setup.mockInput.pressKey("g", { ctrl: true })
@@ -810,7 +811,7 @@ test("keeps assistant footer metrics current after prepend, same-length refresh,
   )
   expect(prepended).toContain("Current input")
   expect(prepended).toContain("Original answer")
-  expect(prepended).toContain("4.0s \u00b7 40.0 tok/s \u00b7 500ms ttft")
+  expect(prepended).toContain("4.0s \u00b7 40.0 tok/s")
 
   // Refresh the latest page: length stays four, but the retained assistant moves from index three to one.
   refresh = true
@@ -826,7 +827,7 @@ test("keeps assistant footer metrics current after prepend, same-length refresh,
   )
   expect(refreshed).toContain("Current input")
   expect(refreshed).toContain("Original answer")
-  expect(refreshed).toContain("4.0s \u00b7 40.0 tok/s \u00b7 500ms ttft")
+  expect(refreshed).toContain("4.0s \u00b7 40.0 tok/s")
   expect(refreshed).toContain("3.0s \u00b7 50.0 tok/s")
 
   setup.events.emit({
@@ -841,7 +842,7 @@ test("keeps assistant footer metrics current after prepend, same-length refresh,
     (frame) => frame.includes("Original answer") && !frame.includes("Later input") && !frame.includes("Later answer"),
   )
   expect(reverted).toContain("Current input")
-  expect(reverted).toContain("4.0s \u00b7 40.0 tok/s \u00b7 500ms ttft")
+  expect(reverted).toContain("4.0s \u00b7 40.0 tok/s")
   expect(reverted).not.toContain("50.0 tok/s")
 })
 
@@ -2844,4 +2845,3 @@ test.each([44, 100])(
   },
   15_000,
 )
-
@@ -22,7 +22,6 @@ test("measures decode throughput from the first streamed token", () => {
 
   // 65 tokens over the 5s the provider spent streaming them, not the 10s that include waiting for the first.
   expect(SessionTurn.tokensPerSecond(messages, final)).toBe(13)
-  expect(SessionTurn.timeToFirstToken(messages, final)).toBe(2_000)
 })
 
 test("falls back to the request window for steps recorded without a first token", () => {
@@ -32,8 +31,6 @@ test("falls back to the request window for steps recorded without a first token"
 
   // The old step counts its 4s request window, the new one its 3s decode window.
   expect(SessionTurn.tokensPerSecond(messages, final)).toBe(65 / 7)
-  // Only the step that recorded a first token can report how long it waited for one.
-  expect(SessionTurn.timeToFirstToken(messages, final)).toBe(1_000)
 })
 
 test("measures request throughput including reasoning without tool time", () => {
@@ -42,7 +39,6 @@ test("measures request throughput including reasoning without tool time", () =>
   const messages = [input("user-1", 1_000), first, final]
 
   expect(SessionTurn.tokensPerSecond(messages, final)).toBe(6.5)
-  expect(SessionTurn.timeToFirstToken(messages, final)).toBeUndefined()
   first.time.streamed = undefined
   expect(SessionTurn.tokensPerSecond(messages, final)).toBeUndefined()
 })
 
@@ -55,16 +55,6 @@ export function tokensPerSecond(messages: readonly Message[], message: Message,
   return output / (total / 1_000)
 }
 
-/**
- * Milliseconds from the dispatch of the turn's first token-producing step to its first streamed token, or undefined
- * until a step records one.
- */
-export function timeToFirstToken(messages: readonly Message[], message: Message, options: Options = {}) {
-  const step = turnSteps(messages, message, options).find((item) => item.time.firstToken !== undefined)
-  if (step?.time.firstToken === undefined) return
-  return Math.max(0, step.time.firstToken - step.time.created)
-}
-
 export function legacy(messages: readonly Message[]) {
   return !messages.some((message) => message.type === "idle")
 }