diff --git a/packages/core/src/aisdk.ts b/packages/core/src/aisdk.ts
index 6d749b8526..e22a909f0a 100644
--- a/packages/core/src/aisdk.ts
+++ b/packages/core/src/aisdk.ts
@@ -129,20 +129,6 @@ function prepareOptions(model: ModelV2.Info, pkg: string) {
     if (abortSignals.length === 1) opts.signal = abortSignals[0]
     if (abortSignals.length > 1) opts.signal = AbortSignal.any(abortSignals)
 
-    if (
-      (pkg === "@ai-sdk/openai" || pkg === "@ai-sdk/azure" || pkg === "@ai-sdk/amazon-bedrock/mantle") &&
-      opts.body &&
-      opts.method === "POST"
-    ) {
-      const body = JSON.parse(opts.body as string)
-      if (body.store !== true && Array.isArray(body.input)) {
-        for (const item of body.input) {
-          if ("id" in item) delete item.id
-        }
-        opts.body = JSON.stringify(body)
-      }
-    }
-
     if (typeof opts.body === "string" && model.body !== undefined) {
       const decoded = Option.getOrUndefined(Schema.decodeUnknownOption(Schema.UnknownFromJsonString)(opts.body))
       if (Schema.is(Schema.Record(Schema.String, Schema.Json))(decoded)) {
@@ -337,7 +323,7 @@ function modelFromLanguage(info: ModelV2.Info, language: LanguageModelV3) {
     },
     body: {
       schema: Schema.Unknown,
-      from: (request) => Effect.succeed(callOptions(request)),
+      from: (request) => Effect.succeed(callOptions(request, info)),
     },
     with: () => route,
     model: (input) => Model.make({ ...input, provider: "provider" in input ? input.provider : info.providerID, route }),
@@ -411,8 +397,11 @@ function mapBodyToProviderOptions(model: ModelV2.Info, packageName: string) {
   }
 }
 
-function callOptions(request: LLMRequest): LanguageModelV3CallOptions {
-  return {
+function callOptions(request: LLMRequest, model: ModelV2.Info): LanguageModelV3CallOptions {
+  const packageName = ProviderV2.isAISDK(model.package) ? ProviderV2.packageName(model.package) : undefined
+  const optionKey = providerOptionKey(packageName, model.providerID)
+  const store = request.providerOptions?.[optionKey]?.store
+  const options: LanguageModelV3CallOptions = {
     prompt: prompt(request),
     maxOutputTokens: request.generation?.maxTokens ?? request.model.route.defaults.limits?.output,
     temperature: request.generation?.temperature,
@@ -428,6 +417,30 @@ function callOptions(request: LLMRequest): LanguageModelV3CallOptions {
     headers: request.http?.headers,
     providerOptions: providerOptions(request.providerOptions),
   }
+  if (
+    store !== true &&
+    packageName !== undefined &&
+    ["@ai-sdk/openai", "@ai-sdk/azure", "@ai-sdk/amazon-bedrock/mantle", "@ai-sdk/github-copilot"].includes(
+      packageName,
+    )
+  ) {
+    options.prompt = options.prompt.map((message) => {
+      if (!Array.isArray(message.content)) return message
+      return {
+        ...message,
+        content: message.content.map((part) => {
+          if (!("providerOptions" in part)) return part
+          const providerOptions = part.providerOptions
+          const metadata = providerOptions?.[optionKey]
+          if (metadata === undefined || !("itemId" in metadata)) return part
+          const next = { ...metadata }
+          delete next.itemId
+          return { ...part, providerOptions: { ...providerOptions, [optionKey]: next } }
+        }),
+      } as LanguageModelV3Message
+    })
+  }
+  return options
 }
 
 function prompt(request: LLMRequest): LanguageModelV3Prompt {
diff --git a/packages/core/src/github-copilot/responses/convert-to-openai-responses-input.ts b/packages/core/src/github-copilot/responses/convert-to-openai-responses-input.ts
index 1e4f86d933..80aea89728 100644
--- a/packages/core/src/github-copilot/responses/convert-to-openai-responses-input.ts
+++ b/packages/core/src/github-copilot/responses/convert-to-openai-responses-input.ts
@@ -237,10 +237,19 @@ export async function convertToOpenAIResponsesInput({
                   }
                 }
               } else {
-                warnings.push({
-                  type: "other",
-                  message: `Non-OpenAI reasoning parts are not supported. Skipping reasoning part: ${JSON.stringify(part)}.`,
-                })
+                const encryptedContent = providerOptions?.reasoningEncryptedContent
+                if (encryptedContent != null) {
+                  input.push({
+                    type: "reasoning",
+                    encrypted_content: encryptedContent,
+                    summary: part.text.length > 0 ? [{ type: "summary_text", text: part.text }] : [],
+                  })
+                } else {
+                  warnings.push({
+                    type: "other",
+                    message: `Non-OpenAI reasoning parts are not supported. Skipping reasoning part: ${JSON.stringify(part)}.`,
+                  })
+                }
               }
               break
             }
diff --git a/packages/core/src/github-copilot/responses/openai-responses-api-types.ts b/packages/core/src/github-copilot/responses/openai-responses-api-types.ts
index dfdd066750..56d44de884 100644
--- a/packages/core/src/github-copilot/responses/openai-responses-api-types.ts
+++ b/packages/core/src/github-copilot/responses/openai-responses-api-types.ts
@@ -205,7 +205,7 @@ export type OpenAIResponsesTool =
 
 export type OpenAIResponsesReasoning = {
   type: "reasoning"
-  id: string
+  id?: string
   encrypted_content?: string | null
   summary: Array<{
     type: "summary_text"
diff --git a/packages/core/test/aisdk.test.ts b/packages/core/test/aisdk.test.ts
index 0bc16e8f59..d15c1c9fc4 100644
--- a/packages/core/test/aisdk.test.ts
+++ b/packages/core/test/aisdk.test.ts
@@ -270,6 +270,140 @@ it.effect("projects replay metadata onto AI SDK prompt parts", () =>
   }),
 )
 
+it.effect("strips stale Azure item IDs before stateless AI SDK serialization", () =>
+  Effect.gen(function* () {
+    const aisdk = yield* AISDK.Service
+    yield* aisdk.hook.sdk((event) => {
+      event.sdk = { languageModel: () => ({ provider: event.model.providerID }) }
+    })
+
+    const messages = [
+      Message.assistant([
+        {
+          type: "reasoning" as const,
+          text: "Think",
+          providerMetadata: {
+            azure: { itemId: "rs_123", reasoningEncryptedContent: "encrypted" },
+          },
+        },
+        {
+          type: "tool-call" as const,
+          id: "call_1",
+          name: "lookup",
+          input: { query: "Effect" },
+          providerMetadata: { azure: { itemId: "fc_123", otherOption: "value" } },
+        },
+      ]),
+    ]
+    const resolved = yield* aisdk.model(model("@ai-sdk/azure", { store: false }))
+    const prepared = yield* LLMClient.prepare<LanguageModelV3CallOptions>(
+      LLM.request({
+        model: resolved,
+        messages,
+      }),
+    )
+
+    expect(prepared.body.prompt).toEqual([
+      {
+        role: "assistant",
+        content: [
+          {
+            type: "reasoning",
+            text: "Think",
+            providerOptions: { azure: { reasoningEncryptedContent: "encrypted" } },
+          },
+          {
+            type: "tool-call",
+            toolCallId: "call_1",
+            toolName: "lookup",
+            input: { query: "Effect" },
+            providerExecuted: undefined,
+            providerOptions: { azure: { otherOption: "value" } },
+          },
+        ],
+      },
+    ])
+
+    const stateful = yield* aisdk.model(model("@ai-sdk/azure", { store: true }))
+    const statefulPrepared = yield* LLMClient.prepare<LanguageModelV3CallOptions>(
+      LLM.request({ model: stateful, messages }),
+    )
+    expect(statefulPrepared.body.prompt[0]).toMatchObject({
+      role: "assistant",
+      content: [
+        { providerOptions: { azure: { itemId: "rs_123" } } },
+        { providerOptions: { azure: { itemId: "fc_123" } } },
+      ],
+    })
+
+    const copilot = yield* aisdk.model(model("@ai-sdk/github-copilot", { store: false }))
+    const copilotPrepared = yield* LLMClient.prepare<LanguageModelV3CallOptions>(
+      LLM.request({
+        model: copilot,
+        messages: [
+          Message.assistant([
+            {
+              type: "reasoning",
+              text: "Think",
+              providerMetadata: {
+                copilot: { itemId: "rs_123", reasoningEncryptedContent: "encrypted" },
+              },
+            },
+          ]),
+        ],
+      }),
+    )
+    expect(copilotPrepared.body.prompt[0]).toMatchObject({
+      role: "assistant",
+      content: [
+        {
+          providerOptions: { copilot: { reasoningEncryptedContent: "encrypted" } },
+        },
+      ],
+    })
+  }),
+)
+
+it.effect("does not corrupt AI SDK item reference payloads", () =>
+  Effect.gen(function* () {
+    const aisdk = yield* AISDK.Service
+    let sent: unknown
+    const server = yield* Effect.acquireRelease(
+      Effect.sync(() =>
+        Bun.serve({
+          port: 0,
+          async fetch(request) {
+            sent = await request.json()
+            return new Response(null, { status: 200 })
+          },
+        }),
+      ),
+      (server) => Effect.sync(() => server.stop(true)),
+    )
+    let wrapped: typeof fetch | undefined
+    yield* aisdk.hook.sdk((event) => {
+      wrapped = event.options.fetch
+      event.sdk = { languageModel: () => ({ provider: event.model.providerID }) }
+    })
+
+    yield* aisdk.language(model("@ai-sdk/azure"))
+    yield* Effect.promise(() =>
+      wrapped!(new URL("responses", server.url), {
+        method: "POST",
+        body: JSON.stringify({
+          store: false,
+          input: [{ type: "item_reference", id: "fc_123" }],
+        }),
+      }),
+    )
+
+    expect(sent).toEqual({
+      store: false,
+      input: [{ type: "item_reference", id: "fc_123" }],
+    })
+  }),
+)
+
 it.effect("emits malformed AI SDK tool input without executing it", () =>
   Effect.gen(function* () {
     const aisdk = yield* AISDK.Service
diff --git a/packages/core/test/github-copilot/openai-responses-language-model.test.ts b/packages/core/test/github-copilot/openai-responses-language-model.test.ts
index ab047d04eb..b7d4d0ffcd 100644
--- a/packages/core/test/github-copilot/openai-responses-language-model.test.ts
+++ b/packages/core/test/github-copilot/openai-responses-language-model.test.ts
@@ -163,7 +163,35 @@ describe("convertToOpenAIResponsesInput", () => {
     ])
   })
 
-  test("drops reasoning items with no copilot itemId and warns, as before", async () => {
+  test("preserves encrypted reasoning with no copilot itemId", async () => {
+    const { input, warnings } = await convertToOpenAIResponsesInput({
+      prompt: [
+        {
+          role: "assistant",
+          content: [
+            {
+              type: "reasoning",
+              text: "thinking...",
+              providerOptions: { copilot: { reasoningEncryptedContent: "enc_1" } },
+            },
+          ],
+        },
+      ],
+      systemMessageMode: "system",
+      store: false,
+    })
+
+    expect(warnings).toEqual([])
+    expect(input).toEqual([
+      {
+        type: "reasoning",
+        encrypted_content: "enc_1",
+        summary: [{ type: "summary_text", text: "thinking..." }],
+      },
+    ])
+  })
+
+  test("drops reasoning with neither a copilot itemId nor encrypted content", async () => {
     const { input, warnings } = await convertToOpenAIResponsesInput({
       prompt: [
         {