diff --git a/backend/internal/handler/gateway_handler.go b/backend/internal/handler/gateway_handler.go
index 82181948..2b3703b4 100644
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -238,6 +238,13 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 		var lastFailoverErr *service.UpstreamFailoverError
 		var forceCacheBilling bool // 粘性会话切换时的缓存计费标记
 
+		// 单账号分组提前设置 SingleAccountRetry 标记，让 Service 层首次 503 就不设模型限流标记。
+		// 避免单账号分组收到 503 (MODEL_CAPACITY_EXHAUSTED) 时设 29s 限流，导致后续请求连续快速失败。
+		if h.gatewayService.IsSingleAntigravityAccountGroup(c.Request.Context(), apiKey.GroupID) {
+			ctx := context.WithValue(c.Request.Context(), ctxkey.SingleAccountRetry, true)
+			c.Request = c.Request.WithContext(ctx)
+		}
+
 		for {
 			selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, reqModel, failedAccountIDs, "") // Gemini 不使用会话限制
 			if err != nil {
@@ -409,6 +416,13 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 	}
 	fallbackUsed := false
 
+	// 单账号分组提前设置 SingleAccountRetry 标记，让 Service 层首次 503 就不设模型限流标记。
+	// 避免单账号分组收到 503 (MODEL_CAPACITY_EXHAUSTED) 时设 29s 限流，导致后续请求连续快速失败。
+	if h.gatewayService.IsSingleAntigravityAccountGroup(c.Request.Context(), currentAPIKey.GroupID) {
+		ctx := context.WithValue(c.Request.Context(), ctxkey.SingleAccountRetry, true)
+		c.Request = c.Request.WithContext(ctx)
+	}
+
 	for {
 		maxAccountSwitches := h.maxAccountSwitches
 		switchCount := 0
diff --git a/backend/internal/handler/gemini_v1beta_handler.go b/backend/internal/handler/gemini_v1beta_handler.go
index 2b67cb1f..f8fb0dcb 100644
--- a/backend/internal/handler/gemini_v1beta_handler.go
+++ b/backend/internal/handler/gemini_v1beta_handler.go
@@ -327,6 +327,13 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
 	var lastFailoverErr *service.UpstreamFailoverError
 	var forceCacheBilling bool // 粘性会话切换时的缓存计费标记
 
+	// 单账号分组提前设置 SingleAccountRetry 标记，让 Service 层首次 503 就不设模型限流标记。
+	// 避免单账号分组收到 503 (MODEL_CAPACITY_EXHAUSTED) 时设 29s 限流，导致后续请求连续快速失败。
+	if h.gatewayService.IsSingleAntigravityAccountGroup(c.Request.Context(), apiKey.GroupID) {
+		ctx := context.WithValue(c.Request.Context(), ctxkey.SingleAccountRetry, true)
+		c.Request = c.Request.WithContext(ctx)
+	}
+
 	for {
 		selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), apiKey.GroupID, sessionKey, modelName, failedAccountIDs, "") // Gemini 不使用会话限制
 		if err != nil {
diff --git a/backend/internal/service/gateway_service.go b/backend/internal/service/gateway_service.go
index 4e723232..2c04ae14 100644
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -1683,6 +1683,17 @@ func (s *GatewayService) listSchedulableAccounts(ctx context.Context, groupID *i
 	return accounts, useMixed, nil
 }
 
+// IsSingleAntigravityAccountGroup 检查指定分组是否只有一个 antigravity 平台的可调度账号。
+// 用于 Handler 层在首次请求时提前设置 SingleAccountRetry context，
+// 避免单账号分组收到 503 时错误地设置模型限流标记导致后续请求连续快速失败。
+func (s *GatewayService) IsSingleAntigravityAccountGroup(ctx context.Context, groupID *int64) bool {
+	accounts, _, err := s.listSchedulableAccounts(ctx, groupID, PlatformAntigravity, true)
+	if err != nil {
+		return false
+	}
+	return len(accounts) == 1
+}
+
 func (s *GatewayService) isAccountAllowedForPlatform(account *Account, platform string, useMixed bool) bool {
 	if account == nil {
 		return false