diff --git a/backend/internal/service/gemini_messages_compat_service.go b/backend/internal/service/gemini_messages_compat_service.go
index aea880c2..cbbf5fcf 100644
--- a/backend/internal/service/gemini_messages_compat_service.go
+++ b/backend/internal/service/gemini_messages_compat_service.go
@@ -2522,9 +2522,13 @@ func extractGeminiUsage(geminiResp map[string]any) *ClaudeUsage {
 	}
 	prompt, _ := asInt(usageMeta["promptTokenCount"])
 	cand, _ := asInt(usageMeta["candidatesTokenCount"])
+	cached, _ := asInt(usageMeta["cachedContentTokenCount"])
+	// 注意：Gemini 的 promptTokenCount 包含 cachedContentTokenCount，
+	// 但 Claude 的 input_tokens 不包含 cache_read_input_tokens，需要减去
 	return &ClaudeUsage{
-		InputTokens:  prompt,
-		OutputTokens: cand,
+		InputTokens:          prompt - cached,
+		OutputTokens:         cand,
+		CacheReadInputTokens: cached,
 	}
 }