diff --git a/packages/core/src/core/openaiContentGenerator/converter.ts b/packages/core/src/core/openaiContentGenerator/converter.ts
index 07a8f1831..184ba5493 100644
--- a/packages/core/src/core/openaiContentGenerator/converter.ts
+++ b/packages/core/src/core/openaiContentGenerator/converter.ts
@@ -696,6 +696,17 @@ export class OpenAIContentConverter {
       parts.push({ text: choice.message.content });
     }
 
+    // Handle reasoning content
+    const message = choice.message as typeof choice.message & {
+      reasoning_content?: string;
+    };
+    if (message.reasoning_content) {
+      parts.push({
+        text: message.reasoning_content,
+        thought: true,
+      } as unknown as Part);
+    }
+
     // Handle tool calls
     if (choice.message.tool_calls) {
       for (const toolCall of choice.message.tool_calls) {
@@ -752,6 +763,8 @@ export class OpenAIContentConverter {
         usage.prompt_tokens_details?.cached_tokens ??
         extendedUsage.cached_tokens ??
         0;
+      const reasoningTokens =
+        usage.completion_tokens_details?.reasoning_tokens || 0;
 
       // If we only have total tokens but no breakdown, estimate the split
       // Typically input is ~70% and output is ~30% for most conversations
@@ -769,6 +782,7 @@ export class OpenAIContentConverter {
         candidatesTokenCount: finalCompletionTokens,
         totalTokenCount: totalTokens,
         cachedContentTokenCount: cachedTokens,
+        thoughtsTokenCount: reasoningTokens,
       };
     }
 
@@ -800,6 +814,17 @@ export class OpenAIContentConverter {
         }
       }
 
+      // Handle reasoning content
+      const delta = choice.delta as typeof choice.delta & {
+        reasoning_content?: string;
+      };
+      if (delta.reasoning_content) {
+        parts.push({
+          text: delta.reasoning_content,
+          thought: true,
+        });
+      }
+
       // Handle tool calls using the streaming parser
       if (choice.delta?.tool_calls) {
         for (const toolCall of choice.delta.tool_calls) {
diff --git a/packages/core/src/core/openaiContentGenerator/pipeline.ts b/packages/core/src/core/openaiContentGenerator/pipeline.ts
index 88ac38f6a..0eab0c2d2 100644
--- a/packages/core/src/core/openaiContentGenerator/pipeline.ts
+++ b/packages/core/src/core/openaiContentGenerator/pipeline.ts
@@ -242,6 +242,30 @@ export class ContentGenerationPipeline {
       baseRequest.stream_options = { include_usage: true };
     }
 
+    // Add thinking options if present
+    if (
+      request.config?.thinkingConfig &&
+      request.config.thinkingConfig.includeThoughts
+    ) {
+      (
+        baseRequest as OpenAI.Chat.ChatCompletionCreateParams & {
+          extra_body?: Record<string, unknown>;
+        }
+      ).extra_body = { enable_thinking: true };
+      (
+        baseRequest as OpenAI.Chat.ChatCompletionCreateParams & {
+          enable_thinking?: boolean;
+        }
+      ).enable_thinking = true;
+      if (request.config.thinkingConfig.thinkingBudget) {
+        (
+          baseRequest as OpenAI.Chat.ChatCompletionCreateParams & {
+            thinking_budget?: number;
+          }
+        ).thinking_budget = request.config.thinkingConfig.thinkingBudget;
+      }
+    }
+
     // Add tools if present
     if (request.config?.tools) {
       baseRequest.tools = await this.converter.convertGeminiToolsToOpenAI(