diff --git a/packages/ai/src/utils/assistant-message-frame.ts b/packages/ai/src/utils/assistant-message-frame.ts
index f314a9f1d..65e23ede5 100644
--- a/packages/ai/src/utils/assistant-message-frame.ts
+++ b/packages/ai/src/utils/assistant-message-frame.ts
@@ -83,6 +83,7 @@ function cloneStartMessage(message: AssistantMessage): AssistantMessage {
 		model: message.model,
 		...(message.responseModel === undefined ? {} : { responseModel: message.responseModel }),
 		...(message.responseId === undefined ? {} : { responseId: message.responseId }),
+		...(message.providerThinkingLevel === undefined ? {} : { providerThinkingLevel: message.providerThinkingLevel }),
 		...(message.diagnostics === undefined ? {} : { diagnostics: structuredClone(message.diagnostics) }),
 		usage: structuredClone(message.usage),
 		stopReason: "pending",
diff --git a/packages/ai/test/assistant-message-frame.test.ts b/packages/ai/test/assistant-message-frame.test.ts
index 0c0564e23..ab7f5f9ac 100644
--- a/packages/ai/test/assistant-message-frame.test.ts
+++ b/packages/ai/test/assistant-message-frame.test.ts
@@ -61,6 +61,16 @@ describe("assistant message frames", () => {
 		]);
 	});
 
+	it("preserves provider thinking level from the stream start", () => {
+		const partial = seed();
+		partial.providerThinkingLevel = "high";
+		const encoder = new AssistantMessageFrameEncoder();
+		const start = frame(encoder, { type: "start", partial });
+
+		expect(start).toMatchObject({ type: "start", partial: { providerThinkingLevel: "high" } });
+		expect(reduceAssistantMessageFrames([start])?.providerThinkingLevel).toBe("high");
+	});
+
 	it("preserves initial and final thinking metadata, including redaction", () => {
 		const partial = seed();
 		const encoder = new AssistantMessageFrameEncoder();