diff --git a/packages/ai/src/protocols/open-responses.ts b/packages/ai/src/protocols/open-responses.ts
index f767fb2e9b75..267e67089392 100644
--- a/packages/ai/src/protocols/open-responses.ts
+++ b/packages/ai/src/protocols/open-responses.ts
@@ -287,6 +287,7 @@ export const Event = Schema.StructWithRest(
     delta: Schema.optional(Schema.String),
     text: Schema.optional(Schema.String),
     item_id: Schema.optional(Schema.String),
+    output_index: Schema.optional(Schema.Number),
     summary_index: Schema.optional(Schema.Number),
     item: Schema.optional(StreamItem),
     response: Schema.optional(
@@ -297,6 +298,7 @@ export const Event = Schema.StructWithRest(
           incomplete_details: optionalNull(Schema.Struct({ reason: Schema.optional(Schema.String) })),
           usage: optionalNull(OpenResponsesUsage),
           error: optionalNull(OpenResponsesErrorPayload),
+          output: Schema.optional(Schema.Array(StreamItem)),
         }),
         [Schema.Record(Schema.String, Schema.Unknown)],
       ),
@@ -342,13 +344,14 @@ export interface ParserState {
   readonly messageItems: ReadonlySet<string>
   readonly messagePhases: Readonly<Record<string, MessagePhase | null>>
   readonly reasoningItems: Readonly<Record<string, ReasoningStreamItem>>
-  readonly store: boolean | undefined
+  readonly reasoningIndexes: Readonly<Record<number, string>>
 }
 
-type ReasoningSummaryStatus = "active" | "can-conclude" | "concluded"
+type ReasoningSummaryStatus = "active" | "concluded"
 
 interface ReasoningStreamItem {
   readonly encryptedContent: string | null | undefined
+  readonly blockIDs?: ReadonlyArray<string>
   // Keyed by the wire protocol's numeric `summary_index`. JS object keys coerce to
   // strings, but typing the map as `Record<number, ...>` documents intent
   // and matches the wire field.
@@ -857,6 +860,10 @@ const onOutputItemAdded = (state: ParserState, event: Event): StepResult => {
           ...state.reasoningItems,
           [item.id]: { encryptedContent: item.encrypted_content, summaryParts: { 0: "active" } },
         },
+        reasoningIndexes:
+          event.output_index === undefined
+            ? state.reasoningIndexes
+            : { ...state.reasoningIndexes, [event.output_index]: item.id },
       },
       events,
     ]
@@ -890,23 +897,11 @@ const onReasoningSummaryPartAdded = (state: ParserState, event: Event): StepResu
   if (event.summary_index === 0) return [state, NO_EVENTS]
 
   const events: LLMEvent[] = []
-  const closed = Object.entries(item.summaryParts)
-    .filter((entry) => entry[1] === "can-conclude")
-    .reduce(
-      (lifecycle, entry) =>
-        Lifecycle.reasoningEnd(
-          lifecycle,
-          events,
-          `${event.item_id}:${entry[0]}`,
-          providerMetadata(state, { itemId: event.item_id }),
-        ),
-      state.lifecycle,
-    )
   return [
     {
       ...state,
       lifecycle: Lifecycle.reasoningStart(
-        closed,
+        state.lifecycle,
         events,
         `${event.item_id}:${event.summary_index}`,
         providerMetadata(state, { itemId: event.item_id, reasoningEncryptedContent: item.encryptedContent ?? null }),
@@ -916,11 +911,7 @@ const onReasoningSummaryPartAdded = (state: ParserState, event: Event): StepResu
         [event.item_id]: {
           ...item,
           summaryParts: {
-            ...Object.fromEntries(
-              Object.entries(item.summaryParts).map((entry) =>
-                entry[1] === "can-conclude" ? [entry[0], "concluded" as const] : entry,
-              ),
-            ),
+            ...item.summaryParts,
             [event.summary_index]: "active",
           },
         },
@@ -938,22 +929,19 @@ const onReasoningSummaryPartDone = (state: ParserState, event: Event): StepResul
   return [
     {
       ...state,
-      lifecycle:
-        state.store !== false
-          ? Lifecycle.reasoningEnd(
-              state.lifecycle,
-              events,
-              `${event.item_id}:${event.summary_index}`,
-              providerMetadata(state, { itemId: event.item_id }),
-            )
-          : state.lifecycle,
+      lifecycle: Lifecycle.reasoningEnd(
+        state.lifecycle,
+        events,
+        `${event.item_id}:${event.summary_index}`,
+        providerMetadata(state, { itemId: event.item_id }),
+      ),
       reasoningItems: {
         ...state.reasoningItems,
         [event.item_id]: {
           ...item,
           summaryParts: {
             ...item.summaryParts,
-            [event.summary_index]: state.store !== false ? "concluded" : "can-conclude",
+            [event.summary_index]: "concluded",
           },
         },
       },
@@ -1039,27 +1027,74 @@ const onOutputItemDone = Effect.fn("OpenResponses.onOutputItemDone")(function* (
   }
 
   if (isReasoningItem(item)) {
+    if (
+      event.output_index !== undefined &&
+      state.reasoningIndexes[event.output_index] !== undefined &&
+      state.reasoningIndexes[event.output_index] !== item.id
+    )
+      return [state, NO_EVENTS] satisfies StepResult
     const events: LLMEvent[] = []
     const metadata = reasoningMetadata(state, item)
     const reasoningItem = state.reasoningItems[item.id]
+    const reasoningIndexes =
+      event.output_index === undefined
+        ? state.reasoningIndexes
+        : Object.fromEntries(
+            Object.entries(state.reasoningIndexes).filter((entry) => entry[0] !== `${event.output_index}`),
+          )
     if (reasoningItem) {
-      const lifecycle = Object.entries(reasoningItem.summaryParts)
-        .filter((entry) => entry[1] === "active" || entry[1] === "can-conclude")
-        .reduce(
-          (lifecycle, entry) => Lifecycle.reasoningEnd(lifecycle, events, `${item.id}:${entry[0]}`, metadata),
-          state.lifecycle,
-        )
-      const { [item.id]: _removed, ...reasoningItems } = state.reasoningItems
-      return [{ ...state, lifecycle, reasoningItems }, events] satisfies StepResult
+      const openParts = Object.entries(reasoningItem.summaryParts).filter((entry) => entry[1] === "active")
+      const lifecycle = openParts.reduce(
+        (lifecycle, entry) => Lifecycle.reasoningEnd(lifecycle, events, `${item.id}:${entry[0]}`, metadata),
+        state.lifecycle,
+      )
+      if (typeof item.encrypted_content === "string" && openParts.length === 0) {
+        const blockID = Object.keys(reasoningItem.summaryParts)
+          .map((index) => `${item.id}:${index}`)
+          .at(-1)
+        if (blockID) events.push(LLMEvent.reasoningMetadata({ id: blockID, providerMetadata: metadata }))
+      }
+      const reasoningItems =
+        typeof item.encrypted_content === "string"
+          ? Object.fromEntries(Object.entries(state.reasoningItems).filter((entry) => entry[0] !== item.id))
+          : {
+              ...state.reasoningItems,
+              [item.id]: {
+                ...reasoningItem,
+                encryptedContent: item.encrypted_content,
+                summaryParts: Object.fromEntries(
+                  Object.keys(reasoningItem.summaryParts).map((index) => [index, "concluded" as const]),
+                ),
+              },
+            }
+      return [{ ...state, lifecycle, reasoningItems, reasoningIndexes }, events] satisfies StepResult
     }
     if (!state.lifecycle.reasoning.has(item.id)) {
       const lifecycle = Lifecycle.stepStart(state.lifecycle, events)
       events.push(LLMEvent.reasoningStart({ id: item.id, providerMetadata: metadata }))
       events.push(LLMEvent.reasoningEnd({ id: item.id, providerMetadata: metadata }))
-      return [{ ...state, lifecycle }, events] satisfies StepResult
+      return [
+        {
+          ...state,
+          lifecycle,
+          reasoningIndexes,
+          reasoningItems:
+            typeof item.encrypted_content === "string"
+              ? state.reasoningItems
+              : {
+                  ...state.reasoningItems,
+                  [item.id]: { encryptedContent: item.encrypted_content, blockIDs: [item.id], summaryParts: {} },
+                },
+        },
+        events,
+      ] satisfies StepResult
     }
     return [
-      { ...state, lifecycle: Lifecycle.reasoningEnd(state.lifecycle, events, item.id, metadata) },
+      {
+        ...state,
+        lifecycle: Lifecycle.reasoningEnd(state.lifecycle, events, item.id, metadata),
+        reasoningIndexes,
+      },
       events,
     ] satisfies StepResult
   }
@@ -1077,7 +1112,27 @@ const onResponseFinish = Effect.fn("OpenResponses.onResponseFinish")(function* (
   const hasFunctionCall =
     pending.events.some((event) => LLMEvent.is.toolCall(event) || LLMEvent.is.toolInputError(event)) ||
     state.hasFunctionCall
-  const lifecycle = Lifecycle.finish(state.lifecycle, events, {
+  const terminalReasoning = new Map(
+    (event.response?.output ?? []).filter(isReasoningItem).map((item) => [item.id, item]),
+  )
+  const reasoningLifecycle = Object.entries(state.reasoningItems).reduce((lifecycle, [id, item]) => {
+    const terminal = terminalReasoning.get(id)
+    if (!terminal) return lifecycle
+    const metadata = providerMetadata(state, {
+      itemId: id,
+      reasoningEncryptedContent: terminal.encrypted_content ?? null,
+    })
+    const blockID =
+      item.blockIDs?.at(-1) ??
+      Object.keys(item.summaryParts)
+        .map((index) => `${id}:${index}`)
+        .at(-1)
+    if (!blockID) return lifecycle
+    if (lifecycle.reasoning.has(blockID)) return Lifecycle.reasoningEnd(lifecycle, events, blockID, metadata)
+    events.push(LLMEvent.reasoningMetadata({ id: blockID, providerMetadata: metadata }))
+    return lifecycle
+  }, state.lifecycle)
+  const lifecycle = Lifecycle.finish(reasoningLifecycle, events, {
     reason: {
       normalized: mapFinishReason(event, hasFunctionCall),
       raw: event.response?.incomplete_details?.reason,
@@ -1091,7 +1146,10 @@ const onResponseFinish = Effect.fn("OpenResponses.onResponseFinish")(function* (
           })
         : undefined,
   })
-  return [{ ...state, lifecycle, hasFunctionCall, tools: pending.tools }, events] satisfies StepResult
+  return [
+    { ...state, lifecycle, hasFunctionCall, tools: pending.tools, reasoningItems: {}, reasoningIndexes: {} },
+    events,
+  ] satisfies StepResult
 })
 
 // Build the prettiest summary available from whatever the provider supplied.
@@ -1210,7 +1268,7 @@ export const initial = (request: LLMRequest, extension: Extension = BASE): Parse
   messageItems: new Set<string>(),
   messagePhases: {},
   reasoningItems: {},
-  store: OpenResponsesOptions.resolve(request).store,
+  reasoningIndexes: {},
 })
 
 export const protocol = Protocol.make({
diff --git a/packages/ai/src/schema/events.ts b/packages/ai/src/schema/events.ts
index f0857840d41e..00c0e1d9333d 100644
--- a/packages/ai/src/schema/events.ts
+++ b/packages/ai/src/schema/events.ts
@@ -138,6 +138,13 @@ export const ReasoningEnd = Schema.Struct({
 }).annotate({ identifier: "LLM.Event.ReasoningEnd" })
 export type ReasoningEnd = Schema.Schema.Type<typeof ReasoningEnd>
 
+export const ReasoningMetadata = Schema.Struct({
+  type: Schema.tag("reasoning-metadata"),
+  id: ContentBlockID,
+  providerMetadata: ProviderMetadata,
+}).annotate({ identifier: "LLM.Event.ReasoningMetadata" })
+export type ReasoningMetadata = Schema.Schema.Type<typeof ReasoningMetadata>
+
 export const ToolInputStart = Schema.Struct({
   type: Schema.tag("tool-input-start"),
   id: ToolCallID,
@@ -242,6 +249,7 @@ const llmEventTagged = Schema.Union([
   ReasoningStart,
   ReasoningDelta,
   ReasoningEnd,
+  ReasoningMetadata,
   ToolInputStart,
   ToolInputDelta,
   ToolInputEnd,
@@ -278,6 +286,8 @@ export const LLMEvent = Object.assign(llmEventTagged, {
     ReasoningDelta.make({ ...input, id: contentBlockID(input.id) }),
   reasoningEnd: (input: WithID<ReasoningEnd, ContentBlockID>) =>
     ReasoningEnd.make({ ...input, id: contentBlockID(input.id) }),
+  reasoningMetadata: (input: WithID<ReasoningMetadata, ContentBlockID>) =>
+    ReasoningMetadata.make({ ...input, id: contentBlockID(input.id) }),
   toolInputStart: (input: WithID<ToolInputStart, ToolCallID>) =>
     ToolInputStart.make({ ...input, id: toolCallID(input.id) }),
   toolInputDelta: (input: WithID<ToolInputDelta, ToolCallID>) =>
@@ -312,6 +322,7 @@ export const LLMEvent = Object.assign(llmEventTagged, {
     reasoningStart: llmEventTagged.guards["reasoning-start"],
     reasoningDelta: llmEventTagged.guards["reasoning-delta"],
     reasoningEnd: llmEventTagged.guards["reasoning-end"],
+    reasoningMetadata: llmEventTagged.guards["reasoning-metadata"],
     toolInputStart: llmEventTagged.guards["tool-input-start"],
     toolInputDelta: llmEventTagged.guards["tool-input-delta"],
     toolInputEnd: llmEventTagged.guards["tool-input-end"],
@@ -483,6 +494,18 @@ const reduceReasoningEnd = (state: ResponseState, event: ReasoningEnd): Response
   }
 }
 
+const reduceReasoningMetadata = (state: ResponseState, event: ReasoningMetadata): ResponseState => {
+  const current = state.reasoningParts[event.id]
+  if (!current) return state
+  return {
+    ...replaceContent(state, current.contentIndex, reasoningContent(current.text, event.providerMetadata)),
+    reasoningParts: {
+      ...state.reasoningParts,
+      [event.id]: { ...current, providerMetadata: event.providerMetadata },
+    },
+  }
+}
+
 const reduceToolInputStart = (state: ResponseState, event: ToolInputStart): ResponseState => ({
   ...state,
   toolInputs: {
@@ -552,6 +575,8 @@ const reduceResponseState = (state: ResponseState, event: LLMEvent): ResponseSta
       return reduceReasoningDelta(next, event)
     case "reasoning-end":
       return reduceReasoningEnd(next, event)
+    case "reasoning-metadata":
+      return reduceReasoningMetadata(next, event)
     case "tool-input-start":
       return reduceToolInputStart(next, event)
     case "tool-input-delta":
diff --git a/packages/ai/test/lib/tool-runtime.ts b/packages/ai/test/lib/tool-runtime.ts
index 629720cd191a..36bad625ea56 100644
--- a/packages/ai/test/lib/tool-runtime.ts
+++ b/packages/ai/test/lib/tool-runtime.ts
@@ -82,16 +82,25 @@ const indexStep = (event: LLMEvent, index: number): LLMEvent => {
 
 const stepState = (events: ReadonlyArray<LLMEvent>) => {
   const assistantContent: ContentPart[] = []
+  const reasoningIndexes = new Map<string, number>()
   const toolCalls: ToolCallPart[] = []
   let reason: Extract<LLMEvent, { type: "finish" }>["reason"] = { normalized: "unknown" }
   let usage: Usage | undefined
   let providerMetadata: ProviderMetadata | undefined
 
   for (const event of events) {
-    if (event.type === "text-delta" || event.type === "reasoning-delta") {
-      appendText(assistantContent, event.type === "text-delta" ? "text" : "reasoning", event.text)
-    } else if (event.type === "text-end" || event.type === "reasoning-end") {
-      appendText(assistantContent, event.type === "text-end" ? "text" : "reasoning", "", event.providerMetadata)
+    if (event.type === "text-delta") {
+      appendText(assistantContent, "text", event.text)
+    } else if (event.type === "reasoning-delta") {
+      appendReasoning(assistantContent, reasoningIndexes, event.id, event.text, event.providerMetadata)
+    } else if (event.type === "text-end") {
+      appendText(assistantContent, "text", "", event.providerMetadata)
+    } else if (event.type === "reasoning-end") {
+      appendReasoning(assistantContent, reasoningIndexes, event.id, "", event.providerMetadata)
+    } else if (event.type === "reasoning-metadata") {
+      const index = reasoningIndexes.get(event.id)
+      const reasoning = index === undefined ? undefined : assistantContent[index]
+      if (reasoning?.type === "reasoning") reasoning.providerMetadata = event.providerMetadata
     } else if (event.type === "tool-call") {
       assistantContent.push(event)
       if (!event.providerExecuted) toolCalls.push(event)
@@ -114,6 +123,28 @@ const stepState = (events: ReadonlyArray<LLMEvent>) => {
   return { assistantContent, toolCalls, reason, usage, providerMetadata }
 }
 
+const appendReasoning = (
+  content: ContentPart[],
+  indexes: Map<string, number>,
+  id: string,
+  text: string,
+  providerMetadata?: ProviderMetadata,
+) => {
+  const index = indexes.get(id)
+  if (index === undefined) {
+    indexes.set(id, content.length)
+    content.push({ type: "reasoning", text, providerMetadata })
+    return
+  }
+  const current = content[index]
+  if (current?.type !== "reasoning") return
+  content[index] = {
+    ...current,
+    text: `${current.text}${text}`,
+    providerMetadata: providerMetadata ?? current.providerMetadata,
+  }
+}
+
 const appendText = (
   content: ContentPart[],
   type: "text" | "reasoning",
diff --git a/packages/ai/test/provider/openai-responses.test.ts b/packages/ai/test/provider/openai-responses.test.ts
index 3587efc1f6bf..6860afe2cc14 100644
--- a/packages/ai/test/provider/openai-responses.test.ts
+++ b/packages/ai/test/provider/openai-responses.test.ts
@@ -2141,6 +2141,281 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
+  it.effect("backfills encrypted reasoning from the terminal response", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.output_item.added", item: { type: "reasoning", id: "rs_1" } },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "rs_1", encrypted_content: null },
+              },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.events.filter((event) => event.type === "reasoning-end")).toEqual([
+        {
+          type: "reasoning-end",
+          id: "rs_1:0",
+          providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: null } },
+        },
+      ])
+      expect(response.events).toContainEqual({
+        type: "reasoning-metadata",
+        id: "rs_1:0",
+        providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+      })
+      expect(response.message.content).toContainEqual({
+        type: "reasoning",
+        text: "thinking",
+        providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+      })
+
+      const prepared = yield* compileRequest(LLM.request({ model, messages: [response.message] }))
+      expect(prepared.body.input).toEqual([
+        {
+          type: "reasoning",
+          id: "rs_1",
+          summary: [{ type: "summary_text", text: "thinking" }],
+          encrypted_content: "terminal-state",
+        },
+      ])
+    }),
+  )
+
+  it.effect("backfills done-only reasoning from the terminal response", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "rs_done", encrypted_content: null },
+              },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_done", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([
+        {
+          type: "reasoning",
+          text: "",
+          providerMetadata: { openai: { itemId: "rs_done", reasoningEncryptedContent: "terminal-state" } },
+        },
+      ])
+      expect(response.events.filter((event) => event.type === "reasoning-end")).toHaveLength(1)
+    }),
+  )
+
+  it.effect("ends reasoning before later tool and text output", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.output_item.added", item: { type: "reasoning", id: "rs_1" } },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "rs_1", encrypted_content: null },
+              },
+              {
+                type: "response.output_item.added",
+                item: { type: "function_call", id: "fc_1", call_id: "call_1", name: "lookup" },
+              },
+              {
+                type: "response.output_item.done",
+                item: {
+                  type: "function_call",
+                  id: "fc_1",
+                  call_id: "call_1",
+                  name: "lookup",
+                  arguments: '{"query":"Effect"}',
+                },
+              },
+              { type: "response.output_item.added", item: { type: "message", id: "msg_1" } },
+              { type: "response.output_text.delta", item_id: "msg_1", delta: "Found it" },
+              { type: "response.output_item.done", item: { type: "message", id: "msg_1" } },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      const reasoningEnd = response.events.findIndex((event) => event.type === "reasoning-end")
+      expect(reasoningEnd).toBeLessThan(response.events.findIndex((event) => event.type === "tool-input-start"))
+      expect(reasoningEnd).toBeLessThan(response.events.findIndex((event) => event.type === "text-start"))
+      expect(response.events.filter((event) => event.type === "reasoning-end")).toHaveLength(1)
+      expect(response.message.content[0]).toEqual({
+        type: "reasoning",
+        text: "thinking",
+        providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+      })
+    }),
+  )
+
+  it.effect("uses terminal reasoning instead of output_item.added encryption", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.added",
+                item: { type: "reasoning", id: "rs_1", encrypted_content: "provisional-state" },
+              },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([
+        {
+          type: "reasoning",
+          text: "thinking",
+          providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+        },
+      ])
+    }),
+  )
+
+  it.effect("backfills reasoning from an incomplete terminal response", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.output_item.added", item: { type: "reasoning", id: "rs_1" } },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "rs_1", encrypted_content: null },
+              },
+              {
+                type: "response.incomplete",
+                response: {
+                  id: "resp_1",
+                  incomplete_details: { reason: "max_output_tokens" },
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.finishReason).toEqual({ normalized: "length", raw: "max_output_tokens" })
+      expect(response.message.content[0]).toMatchObject({
+        providerMetadata: { openai: { reasoningEncryptedContent: "terminal-state" } },
+      })
+    }),
+  )
+
+  it.effect("does not backfill reasoning from a mismatched terminal item", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              { type: "response.output_item.added", item: { type: "reasoning", id: "rs_1" } },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.output_item.done",
+                item: { type: "reasoning", id: "rs_1", encrypted_content: null },
+              },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_other", encrypted_content: "other-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.events.some((event) => event.type === "reasoning-metadata")).toBe(false)
+      expect(response.message.content[0]).toMatchObject({
+        providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: null } },
+      })
+    }),
+  )
+
+  it.effect("ignores a reasoning done item that mismatches its output index", () =>
+    Effect.gen(function* () {
+      const response = yield* LLMClient.generate(request).pipe(
+        Effect.provide(
+          fixedResponse(
+            sseEvents(
+              {
+                type: "response.output_item.added",
+                output_index: 0,
+                item: { type: "reasoning", id: "rs_1" },
+              },
+              { type: "response.reasoning_summary_text.delta", item_id: "rs_1", delta: "thinking" },
+              {
+                type: "response.output_item.done",
+                output_index: 0,
+                item: { type: "reasoning", id: "rs_other", encrypted_content: "wrong-state" },
+              },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
+            ),
+          ),
+        ),
+      )
+
+      expect(response.message.content).toEqual([
+        {
+          type: "reasoning",
+          text: "thinking",
+          providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+        },
+      ])
+    }),
+  )
+
   it.effect("streams each reasoning summary part as a separate block", () =>
     Effect.gen(function* () {
       const response = yield* LLMClient.generate(
@@ -2188,6 +2463,11 @@ describe("OpenAI Responses route", () => {
         {
           type: "reasoning-end",
           id: "rs_1:1",
+          providerMetadata: { openai: { itemId: "rs_1" } },
+        },
+        {
+          type: "reasoning-metadata",
+          id: "rs_1:1",
           providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "encrypted-state" } },
         },
         { type: "step-finish", index: 0, reason: { normalized: "stop", raw: undefined } },
@@ -2196,7 +2476,7 @@ describe("OpenAI Responses route", () => {
     }),
   )
 
-  it.effect("closes reasoning summary parts when storage is not disabled", () =>
+  it.effect("backfills terminal reasoning when storage is enabled", () =>
     Effect.gen(function* () {
       const response = yield* LLMClient.generate(LLMRequest.update(request, { providerOptions: { store: true } })).pipe(
         Effect.provide(
@@ -2216,7 +2496,13 @@ describe("OpenAI Responses route", () => {
                 type: "response.output_item.done",
                 item: { type: "reasoning", id: "rs_1", encrypted_content: null },
               },
-              { type: "response.completed", response: { id: "resp_1" } },
+              {
+                type: "response.completed",
+                response: {
+                  id: "resp_1",
+                  output: [{ type: "reasoning", id: "rs_1", encrypted_content: "terminal-state" }],
+                },
+              },
             ),
           ),
         ),
@@ -2224,8 +2510,17 @@ describe("OpenAI Responses route", () => {
 
       expect(response.events.filter((event) => event.type === "reasoning-end")).toEqual([
         { type: "reasoning-end", id: "rs_1:0", providerMetadata: { openai: { itemId: "rs_1" } } },
-        { type: "reasoning-end", id: "rs_1:1", providerMetadata: { openai: { itemId: "rs_1" } } },
+        {
+          type: "reasoning-end",
+          id: "rs_1:1",
+          providerMetadata: { openai: { itemId: "rs_1" } },
+        },
       ])
+      expect(response.events).toContainEqual({
+        type: "reasoning-metadata",
+        id: "rs_1:1",
+        providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+      })
     }),
   )
 
diff --git a/packages/client/src/effect/api/api.ts b/packages/client/src/effect/api/api.ts
index a2c002a0f1b6..c868c787fb72 100644
--- a/packages/client/src/effect/api/api.ts
+++ b/packages/client/src/effect/api/api.ts
@@ -709,6 +709,20 @@ export type SessionLogOutput =
             readonly state?: SessionMessage.ProviderState | undefined
           }
         }
+      | {
+          readonly id: Event.ID
+          readonly created: number
+          readonly metadata?: { readonly [x: string]: unknown } | undefined
+          readonly type: "session.reasoning.state.updated"
+          readonly durable: { readonly aggregateID: string; readonly seq: Event.Seq; readonly version: Event.Version }
+          readonly location?: Location.Ref | undefined
+          readonly data: {
+            readonly sessionID: Session.ID
+            readonly assistantMessageID: SessionMessage.ID
+            readonly ordinal: number
+            readonly state: SessionMessage.ProviderState
+          }
+        }
       | {
           readonly id: Event.ID
           readonly created: number
diff --git a/packages/client/src/promise/generated/types.ts b/packages/client/src/promise/generated/types.ts
index 2af0d4d96f2c..3913cbe72fa3 100644
--- a/packages/client/src/promise/generated/types.ts
+++ b/packages/client/src/promise/generated/types.ts
@@ -1263,6 +1263,16 @@ export type SessionToolCalled = {
   }
 }
 
+export type SessionReasoningStateUpdated = {
+  id: string
+  created: number
+  metadata?: { [x: string]: any }
+  type: "session.reasoning.state.updated"
+  durable: { aggregateID: string; seq: number; version: 1 }
+  location?: LocationRef
+  data: { sessionID: string; assistantMessageID: string; ordinal: number; state: SessionMessageProviderState1 }
+}
+
 export type ToolContent1 = ToolTextContent | ToolFileContent1
 
 export type ModelCompatibility = {
@@ -1975,6 +1985,7 @@ export type SessionEventDurable =
   | SessionTextEnded
   | SessionReasoningStarted
   | SessionReasoningEnded
+  | SessionReasoningStateUpdated
   | SessionToolInputStarted
   | SessionToolInputEnded
   | SessionToolCalled
@@ -2069,6 +2080,7 @@ export type V2Event =
   | SessionReasoningStarted
   | SessionReasoningDelta
   | SessionReasoningEnded
+  | SessionReasoningStateUpdated
   | SessionToolInputStarted
   | SessionToolInputDelta
   | SessionToolInputEnded
diff --git a/packages/client/src/solid/data.ts b/packages/client/src/solid/data.ts
index 2c3f35dc8200..878c22786be4 100644
--- a/packages/client/src/solid/data.ts
+++ b/packages/client/src/solid/data.ts
@@ -904,6 +904,14 @@ export function createData(config: CreateDataInput) {
           }
         })
         return
+      case "session.reasoning.state.updated":
+        message.update(event.data.sessionID, (draft, index) => {
+          const match = message
+            .assistant(draft, index, event.data.assistantMessageID)
+            ?.content.filter((item) => item.type === "reasoning")[event.data.ordinal]
+          if (match?.type === "reasoning") match.state = event.data.state
+        })
+        return
       case "session.retry.scheduled":
         message.update(event.data.sessionID, (draft, index) => {
           const currentAssistant = message.assistant(draft, index, event.data.assistantMessageID)
diff --git a/packages/core/src/session/message-updater.ts b/packages/core/src/session/message-updater.ts
index f3df3a8151a5..a85c93462fd0 100644
--- a/packages/core/src/session/message-updater.ts
+++ b/packages/core/src/session/message-updater.ts
@@ -373,6 +373,14 @@ export function update(adapter: Adapter, event: SessionEvent.DurableEvent) {
           }
         })
       },
+      "session.reasoning.state.updated": (event) => {
+        return updateOwnedAssistant(event.data.assistantMessageID, (draft) => {
+          const match = draft.content.filter((item): item is DraftReasoning => item.type === "reasoning")[
+            event.data.ordinal
+          ]
+          if (match) match.state = event.data.state
+        })
+      },
       "session.retry.scheduled": (event) => {
         return updateOwnedAssistant(event.data.assistantMessageID, (draft) => {
           draft.retry = {
diff --git a/packages/core/src/session/projector.ts b/packages/core/src/session/projector.ts
index 164db2d508c0..77fd7cfd076d 100644
--- a/packages/core/src/session/projector.ts
+++ b/packages/core/src/session/projector.ts
@@ -661,6 +661,7 @@ const layer = Layer.effectDiscard(
     yield* bus.project(SessionEvent.Tool.Failed, (event) => run(db, event))
     yield* bus.project(SessionEvent.Reasoning.Started, (event) => run(db, event))
     yield* bus.project(SessionEvent.Reasoning.Ended, (event) => run(db, event))
+    yield* bus.project(SessionEvent.Reasoning.StateUpdated, (event) => run(db, event))
     yield* bus.project(SessionEvent.RetryScheduled, (event) => run(db, event))
     yield* bus.project(SessionEvent.Compaction.Started, (event) => run(db, event))
     yield* bus.project(SessionEvent.Compaction.Ended, (event) =>
diff --git a/packages/core/src/session/runner/publish-llm-event.ts b/packages/core/src/session/runner/publish-llm-event.ts
index 8fa3e70da077..1c3118fac5bc 100644
--- a/packages/core/src/session/runner/publish-llm-event.ts
+++ b/packages/core/src/session/runner/publish-llm-event.ts
@@ -181,7 +181,7 @@ export const createLLMEventPublisher = (bus: Pick<Bus.Interface, "publish">, inp
         state === undefined ? current.state : { ...current.state, ...state },
       )
       chunks.delete(id)
-      return undefined
+      return current.ordinal
     })
     const flush = Effect.fnUntraced(function* () {
       for (const id of Array.from(chunks.keys())) yield* end(id)
@@ -235,6 +235,7 @@ export const createLLMEventPublisher = (bus: Pick<Bus.Interface, "publish">, inp
       }),
     true,
   )
+  const completedReasoning = new Map<string, number>()
   const toolInput = fragments("tool input", (id, value) =>
     Effect.gen(function* () {
       const tool = tools.get(id)
@@ -413,8 +414,21 @@ export const createLLMEventPublisher = (bus: Pick<Bus.Interface, "publish">, inp
         yield* reasoning.append(event.id, event.text, providerState(event.providerMetadata))
         return
       case "reasoning-end":
-        yield* reasoning.end(event.id, providerState(event.providerMetadata))
+        completedReasoning.set(event.id, yield* reasoning.end(event.id, providerState(event.providerMetadata)))
         return
+      case "reasoning-metadata": {
+        const ordinal = completedReasoning.get(event.id)
+        if (ordinal === undefined) return yield* Effect.die(new Error(`Reasoning metadata before end: ${event.id}`))
+        const state = providerState(event.providerMetadata)
+        if (state === undefined) return
+        yield* bus.publish(SessionEvent.Reasoning.StateUpdated, {
+          sessionID: input.sessionID,
+          assistantMessageID: yield* currentAssistantMessageID(),
+          ordinal,
+          state,
+        })
+        return
+      }
       case "tool-input-start":
         outputStarted = true
         yield* startToolInput(event)
diff --git a/packages/core/test/session-runner-tool-events.test.ts b/packages/core/test/session-runner-tool-events.test.ts
index 1ed0f16fac53..fad63eb2db6c 100644
--- a/packages/core/test/session-runner-tool-events.test.ts
+++ b/packages/core/test/session-runner-tool-events.test.ts
@@ -203,6 +203,53 @@ test("reasoning state from start, empty delta, and end is merged", async () => {
   })
 })
 
+test("reasoning metadata updates completed reasoning state", async () => {
+  const { published, publisher } = capture("openai")
+  await Effect.runPromise(
+    Effect.forEach(
+      [
+        LLMEvent.reasoningStart({ id: "reasoning" }),
+        LLMEvent.reasoningEnd({
+          id: "reasoning",
+          providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: null } },
+        }),
+        LLMEvent.reasoningMetadata({
+          id: "reasoning",
+          providerMetadata: { openai: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" } },
+        }),
+      ],
+      publisher.publish,
+      { discard: true },
+    ),
+  )
+
+  expect(published.slice(-2).map((event) => event.type)).toEqual([
+    "session.reasoning.ended.1",
+    "session.reasoning.state.updated.1",
+  ])
+  expect(published.at(-1)?.data).toMatchObject({
+    ordinal: 0,
+    state: { itemId: "rs_1", reasoningEncryptedContent: "terminal-state" },
+  })
+})
+
+test("reasoning metadata ignores unrelated provider state", async () => {
+  const { published, publisher } = capture("openai")
+  await Effect.runPromise(
+    Effect.forEach(
+      [
+        LLMEvent.reasoningStart({ id: "reasoning" }),
+        LLMEvent.reasoningEnd({ id: "reasoning" }),
+        LLMEvent.reasoningMetadata({ id: "reasoning", providerMetadata: { anthropic: { signature: "ignored" } } }),
+      ],
+      publisher.publish,
+      { discard: true },
+    ),
+  )
+
+  expect(published.some((event) => event.type === "session.reasoning.state.updated.1")).toBe(false)
+})
+
 it.effect("batches text deltas and flushes pending text before the terminal event", () =>
   Effect.gen(function* () {
     const { published, publisher } = capture()
diff --git a/packages/core/test/session-runner.test.ts b/packages/core/test/session-runner.test.ts
index 5f706b69b608..8c4a01728818 100644
--- a/packages/core/test/session-runner.test.ts
+++ b/packages/core/test/session-runner.test.ts
@@ -2903,6 +2903,13 @@ describe("SessionRunnerLLM", () => {
           }),
           LLMEvent.reasoningDelta({ id: "reasoning-openai", text: "Encrypted thought" }),
           LLMEvent.reasoningEnd({
+            id: "reasoning-openai",
+            providerMetadata: {
+              openai: { itemId: "rs_1", reasoningEncryptedContent: null },
+              anthropic: { ignored: true },
+            },
+          }),
+          LLMEvent.reasoningMetadata({
             id: "reasoning-openai",
             providerMetadata: {
               openai: { itemId: "rs_1", reasoningEncryptedContent: "encrypted-state" },
diff --git a/packages/schema/src/session-event.ts b/packages/schema/src/session-event.ts
index 6db772c0757b..f35c9a189861 100644
--- a/packages/schema/src/session-event.ts
+++ b/packages/schema/src/session-event.ts
@@ -415,6 +415,18 @@ export namespace Reasoning {
     },
   })
   export type Ended = typeof Ended.Type
+
+  export const StateUpdated = Event.durable({
+    type: "session.reasoning.state.updated",
+    ...options,
+    schema: {
+      ...Base,
+      assistantMessageID: SessionMessage.ID,
+      ordinal: NonNegativeInt,
+      state: SessionMessage.ProviderState,
+    },
+  })
+  export type StateUpdated = typeof StateUpdated.Type
 }
 
 export namespace Tool {
@@ -624,6 +636,7 @@ export const Definitions = Event.inventory(
   Reasoning.Started,
   Reasoning.Delta,
   Reasoning.Ended,
+  Reasoning.StateUpdated,
   Tool.Input.Started,
   Tool.Input.Delta,
   Tool.Input.Ended,
diff --git a/packages/schema/test/event-manifest.test.ts b/packages/schema/test/event-manifest.test.ts
index 556a91e695a6..336a1f34f1fd 100644
--- a/packages/schema/test/event-manifest.test.ts
+++ b/packages/schema/test/event-manifest.test.ts
@@ -111,6 +111,7 @@ describe("public event manifest", () => {
         "session.tool.failed.2",
         "session.reasoning.started.1",
         "session.reasoning.ended.1",
+        "session.reasoning.state.updated.1",
         "session.retry.scheduled.1",
         "session.compaction.started.1",
         "session.compaction.ended.1",