diff --git a/internal/runtime/executor/usage_helpers.go b/internal/runtime/executor/usage_helpers.go
index 7d8d345e..3aa1e7ff 100644
--- a/internal/runtime/executor/usage_helpers.go
+++ b/internal/runtime/executor/usage_helpers.go
@@ -236,11 +236,7 @@ func parseOpenAIStreamUsage(line []byte) (usage.Detail, bool) {
 	return detail, true
 }
 
-func parseOpenAIResponsesUsage(data []byte) usage.Detail {
-	usageNode := gjson.ParseBytes(data).Get("usage")
-	if !usageNode.Exists() {
-		return usage.Detail{}
-	}
+func parseOpenAIResponsesUsageDetail(usageNode gjson.Result) usage.Detail {
 	detail := usage.Detail{
 		InputTokens:  usageNode.Get("input_tokens").Int(),
 		OutputTokens: usageNode.Get("output_tokens").Int(),
@@ -258,6 +254,14 @@ func parseOpenAIResponsesUsage(data []byte) usage.Detail {
 	return detail
 }
 
+func parseOpenAIResponsesUsage(data []byte) usage.Detail {
+	usageNode := gjson.ParseBytes(data).Get("usage")
+	if !usageNode.Exists() {
+		return usage.Detail{}
+	}
+	return parseOpenAIResponsesUsageDetail(usageNode)
+}
+
 func parseOpenAIResponsesStreamUsage(line []byte) (usage.Detail, bool) {
 	payload := jsonPayload(line)
 	if len(payload) == 0 || !gjson.ValidBytes(payload) {
@@ -267,21 +271,7 @@ func parseOpenAIResponsesStreamUsage(line []byte) (usage.Detail, bool) {
 	if !usageNode.Exists() {
 		return usage.Detail{}, false
 	}
-	detail := usage.Detail{
-		InputTokens:  usageNode.Get("input_tokens").Int(),
-		OutputTokens: usageNode.Get("output_tokens").Int(),
-		TotalTokens:  usageNode.Get("total_tokens").Int(),
-	}
-	if detail.TotalTokens == 0 {
-		detail.TotalTokens = detail.InputTokens + detail.OutputTokens
-	}
-	if cached := usageNode.Get("input_tokens_details.cached_tokens"); cached.Exists() {
-		detail.CachedTokens = cached.Int()
-	}
-	if reasoning := usageNode.Get("output_tokens_details.reasoning_tokens"); reasoning.Exists() {
-		detail.ReasoningTokens = reasoning.Int()
-	}
-	return detail, true
+	return parseOpenAIResponsesUsageDetail(usageNode), true
 }
 
 func parseClaudeUsage(data []byte) usage.Detail {
diff --git a/sdk/api/handlers/openai/openai_handlers.go b/sdk/api/handlers/openai/openai_handlers.go
index c8aaba78..f1dd5a07 100644
--- a/sdk/api/handlers/openai/openai_handlers.go
+++ b/sdk/api/handlers/openai/openai_handlers.go
@@ -536,7 +536,12 @@ func (h *OpenAIAPIHandler) handleNonStreamingResponseViaResponses(c *gin.Context
 	}
 	converted := convertResponsesObjectToChatCompletion(cliCtx, modelName, originalChatJSON, rawJSON, resp)
 	if converted == nil {
-		converted = resp
+		h.WriteErrorResponse(c, &interfaces.ErrorMessage{
+			StatusCode: http.StatusInternalServerError,
+			Error:      fmt.Errorf("failed to convert response to chat completion format"),
+		})
+		cliCancel(fmt.Errorf("response conversion failed"))
+		return
 	}
 	_, _ = c.Writer.Write(converted)
 	cliCancel()
diff --git a/sdk/api/handlers/openai/openai_responses_handlers.go b/sdk/api/handlers/openai/openai_responses_handlers.go
index e6c29001..952e44e0 100644
--- a/sdk/api/handlers/openai/openai_responses_handlers.go
+++ b/sdk/api/handlers/openai/openai_responses_handlers.go
@@ -145,8 +145,11 @@ func (h *OpenAIResponsesAPIHandler) handleNonStreamingResponseViaChat(c *gin.Con
 	var param any
 	converted := responsesconverter.ConvertOpenAIChatCompletionsResponseToOpenAIResponsesNonStream(cliCtx, modelName, originalResponsesJSON, originalResponsesJSON, resp, &param)
 	if converted == "" {
-		_, _ = c.Writer.Write(resp)
-		cliCancel()
+		h.WriteErrorResponse(c, &interfaces.ErrorMessage{
+			StatusCode: http.StatusInternalServerError,
+			Error:      fmt.Errorf("failed to convert chat completion response to responses format"),
+		})
+		cliCancel(fmt.Errorf("response conversion failed"))
 		return
 	}
 	_, _ = c.Writer.Write([]byte(converted))