diff --git a/.changeset/smooth-radios-breathe.md b/.changeset/smooth-radios-breathe.md
new file mode 100644
index 000000000000..9f496f3562fa
--- /dev/null
+++ b/.changeset/smooth-radios-breathe.md
@@ -0,0 +1,5 @@
+---
+'@ai-sdk/openai-compatible': patch
+---
+
+support OpenRouter's reasoning field
diff --git a/packages/openai-compatible/src/openai-compatible-chat-language-model.test.ts b/packages/openai-compatible/src/openai-compatible-chat-language-model.test.ts
index 9dfce5e77911..0abb81774005 100644
--- a/packages/openai-compatible/src/openai-compatible-chat-language-model.test.ts
+++ b/packages/openai-compatible/src/openai-compatible-chat-language-model.test.ts
@@ -71,7 +71,8 @@ describe('config', () => {
 describe('doGenerate', () => {
   function prepareJsonResponse({
     content = '',
-    reasoning_content = '',
+    reasoning_content,
+    reasoning,
     tool_calls,
     function_call,
     usage = {
@@ -87,6 +88,7 @@ describe('doGenerate', () => {
   }: {
     content?: string;
     reasoning_content?: string;
+    reasoning?: string;
     tool_calls?: Array<{
       id: string;
       type: 'function';
@@ -133,6 +135,7 @@ describe('doGenerate', () => {
               role: 'assistant',
               content,
               reasoning_content,
+              reasoning,
               tool_calls,
               function_call,
             },
@@ -172,7 +175,7 @@ describe('doGenerate', () => {
     expect(text).toStrictEqual('Hello, World!');
   });
 
-  it('should extract reasoning content', async () => {
+  it('should extract reasoning content when incomming reasoning_content field', async () => {
     prepareJsonResponse({
       content: 'Hello, World!',
       reasoning_content: 'This is the reasoning behind the response',
@@ -190,6 +193,41 @@ describe('doGenerate', () => {
     );
   });
 
+  it('should extract reasoning content when incomming reasoning field', async () => {
+    prepareJsonResponse({
+      content: 'Hello, World!',
+      reasoning: 'This is the reasoning behind the response',
+    });
+
+    const { text, reasoning } = await model.doGenerate({
+      inputFormat: 'prompt',
+      mode: { type: 'regular' },
+      prompt: TEST_PROMPT,
+    });
+
+    expect(text).toStrictEqual('Hello, World!');
+    expect(reasoning).toStrictEqual(
+      'This is the reasoning behind the response',
+    );
+  });
+
+  it('should extract reasoning content when incomming reasoning_content and reasoning fields', async () => {
+    prepareJsonResponse({
+      content: 'Hello, World!',
+      reasoning_content: '123',
+      reasoning: '456',
+    });
+
+    const { text, reasoning } = await model.doGenerate({
+      inputFormat: 'prompt',
+      mode: { type: 'regular' },
+      prompt: TEST_PROMPT,
+    });
+
+    expect(text).toStrictEqual('Hello, World!');
+    expect(reasoning).toStrictEqual('123');
+  });
+
   it('should extract usage', async () => {
     prepareJsonResponse({
       content: '',
@@ -289,7 +327,7 @@ describe('doGenerate', () => {
 
     expect(rawResponse?.headers).toStrictEqual({
       // default headers:
-      'content-length': '335',
+      'content-length': '312',
       'content-type': 'application/json',
 
       // custom header
diff --git a/packages/openai-compatible/src/openai-compatible-chat-language-model.ts b/packages/openai-compatible/src/openai-compatible-chat-language-model.ts
index 20e53eccea9e..740df8dddb39 100644
--- a/packages/openai-compatible/src/openai-compatible-chat-language-model.ts
+++ b/packages/openai-compatible/src/openai-compatible-chat-language-model.ts
@@ -300,7 +300,10 @@ export class OpenAICompatibleChatLanguageModel implements LanguageModelV1 {
 
     return {
       text: choice.message.content ?? undefined,
-      reasoning: choice.message.reasoning_content ?? undefined,
+      reasoning:
+        choice.message.reasoning_content ??
+        choice.message.reasoning ??
+        undefined,
       toolCalls: choice.message.tool_calls?.map(toolCall => ({
         toolCallType: 'function',
         toolCallId: toolCall.id ?? generateId(),
@@ -529,10 +532,10 @@ export class OpenAICompatibleChatLanguageModel implements LanguageModelV1 {
             const delta = choice.delta;
 
             // enqueue reasoning before text deltas:
-            if (delta.reasoning_content != null) {
+            if (delta.reasoning_content != null || delta.reasoning != null) {
               controller.enqueue({
                 type: 'reasoning',
-                textDelta: delta.reasoning_content,
+                textDelta: delta.reasoning_content || delta.reasoning,
               });
             }
 
@@ -730,6 +733,7 @@ const OpenAICompatibleChatResponseSchema = z.object({
         role: z.literal('assistant').nullish(),
         content: z.string().nullish(),
         reasoning_content: z.string().nullish(),
+        reasoning: z.string().nullish(),
         tool_calls: z
           .array(
             z.object({
@@ -766,6 +770,7 @@ const createOpenAICompatibleChatChunkSchema = <ERROR_SCHEMA extends z.ZodType>(
               role: z.enum(['assistant']).nullish(),
               content: z.string().nullish(),
               reasoning_content: z.string().nullish(),
+              reasoning: z.string().nullish(),
               tool_calls: z
                 .array(
                   z.object({