diff --git a/backend/internal/handler/gateway_handler.go b/backend/internal/handler/gateway_handler.go
index ecd7bc7e38..8e1bbd2ae0 100644
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -1440,7 +1440,7 @@ func modelListingSource(platform string, availableModels, fallbackModels []strin
 func defaultCodexModelIDsForPlatform(platform string) []string {
 	switch platform {
 	case service.PlatformDeepseek:
-		return []string{"deepseek-v4-pro", "deepseek-v4-flash"}
+		return []string{"deepseek-v4-pro", "deepseek-v4-flash", "deepseek-flash"}
 	case service.PlatformMiniMax:
 		return []string{"MiniMax-M3", "MiniMax-M2.7", "MiniMax-M2.5"}
 	default:
diff --git a/backend/internal/handler/gateway_models_test.go b/backend/internal/handler/gateway_models_test.go
index 9252851cd6..26c0c72ec4 100644
--- a/backend/internal/handler/gateway_models_test.go
+++ b/backend/internal/handler/gateway_models_test.go
@@ -914,7 +914,7 @@ func TestDefaultModelIDsForPlatform_CNProvidersKeepClaudeDefaults(t *testing.T)
 }
 
 func TestDefaultCodexModelIDsForPlatform_DeepSeekUsesDeepSeekModels(t *testing.T) {
-	require.Equal(t, []string{"deepseek-v4-pro", "deepseek-v4-flash"}, defaultCodexModelIDsForPlatform(service.PlatformDeepseek))
+	require.Equal(t, []string{"deepseek-v4-pro", "deepseek-v4-flash", "deepseek-flash"}, defaultCodexModelIDsForPlatform(service.PlatformDeepseek))
 	require.Equal(t, []string{"MiniMax-M3", "MiniMax-M2.7", "MiniMax-M2.5"}, defaultCodexModelIDsForPlatform(service.PlatformMiniMax))
 	require.Equal(t, defaultModelIDsForPlatform(service.PlatformAnthropic), defaultCodexModelIDsForPlatform(service.PlatformAnthropic))
 }
diff --git a/backend/internal/service/account_stats_pricing_test.go b/backend/internal/service/account_stats_pricing_test.go
index 7d4d8c7d03..e53810b279 100644
--- a/backend/internal/service/account_stats_pricing_test.go
+++ b/backend/internal/service/account_stats_pricing_test.go
@@ -659,7 +659,7 @@ func TestTryModelFilePricing_DeepSeekPeakPricing(t *testing.T) {
 		name                          string
 		input, output, cacheReadPrice float64
 	}{
-		{"deepseek-v4-flash", 2.2e-7, 6.6e-7, 7e-9},
+		{"deepseek-v4-flash", 1.5e-7, 6e-7, 3e-9},
 		{"deepseek-v4-pro", 6.6e-7, 1.98e-6, 2.2e-8},
 	} {
 		for _, usage := range []struct {
@@ -711,7 +711,7 @@ func TestResolveAccountStatsCost_DeepSeekPricingPriority(t *testing.T) {
 		noChannel    bool
 		want         float64
 	}{
-		{name: "catalog", want: 1000 * 2.2e-7 * 2},
+		{name: "catalog", want: 1000 * 1.5e-7 * 2},
 		{name: "custom_rule", customRule: true, want: 1},
 		{name: "custom_rule_before_customer_price", customRule: true, applyPricing: true, want: 1},
 		{name: "customer_price", applyPricing: true, want: 0.75},
diff --git a/backend/internal/service/billing_context_schedule.go b/backend/internal/service/billing_context_schedule.go
index d1f994707c..0be6d7a07a 100644
--- a/backend/internal/service/billing_context_schedule.go
+++ b/backend/internal/service/billing_context_schedule.go
@@ -7,6 +7,8 @@ import (
 	"sort"
 	"strconv"
 	"time"
+
+	"github.com/Wei-Shaw/sub2api/internal/pkg/timezone"
 )
 
 // ContextPricingBasis 阶梯的计价基准。当前只有整单口径；历史上的
@@ -222,7 +224,10 @@ func (s *BillingService) contextPricingBreakpoints(resolver *ModelPricingResolve
 	if pricing == nil {
 		return plan
 	}
-	pricing = s.applyModelSpecificPricingPolicy(model, pricing)
+	// 该路径无既有计费时点（ContextPricingScheduleInput 无时间字段），显式传
+	// 当前时刻；此处 pricing 仅取 LongContextInputThreshold 等时间无关字段，
+	// DeepSeek pro→Flash 切换不影响断点结果。
+	pricing = s.applyModelSpecificPricingPolicyEx(model, pricing, true, timezone.Now())
 	if pricing.LongContextInputThreshold <= 0 {
 		return plan
 	}
diff --git a/backend/internal/service/billing_service.go b/backend/internal/service/billing_service.go
index f09f606eb7..76c1adfc43 100644
--- a/backend/internal/service/billing_service.go
+++ b/backend/internal/service/billing_service.go
@@ -262,14 +262,21 @@ func resolvedChannelTimeMultiplier(resolved *ResolvedPricing, at time.Time) floa
 // sources can price the requested model.
 var ErrModelPricingUnavailable = errors.New("pricing not found")
 
-// ---- DeepSeek 官方低谷价（$/token，2026-08-23 起生效）----
-// Source: https://api-docs.deepseek.com/quick_start/pricing
+// ---- DeepSeek 官方低谷价（$/token）----
+// 2026-09-10 官方公告：DeepSeek-V4.1-Flash（新名 deepseek-flash）大幅降价，
+// Flash 低谷价降为 $0.15/$0.60/$0.003 per MTok（输入缓存未命中/输出/缓存命中）；
+// deepseek-v4-pro 名义价格暂不变，但自北京时间 2026-09-14 12:00（04:00 UTC）起
+// 其请求被上游路由到 V4.1-Flash 并按 Flash 价计费（见 deepseekProBilledAsFlash）。
+// Source: https://api-docs.deepseek.com/news/news260910
+//
+//	https://api-docs.deepseek.com/quick_start/pricing
+//
 // 高峰价 = 2× 低谷价；高峰时段 01:00–04:00 与 06:00–10:00 UTC（仅工作日），
 // 北京时间周六/周日全天低谷。时段判定见 deepseekPeakMultiplierAt。
 const (
-	deepseekFlashOffPeakInputPrice  = 2.2e-7  // $0.22 per MTok (cache miss)
-	deepseekFlashOffPeakOutputPrice = 6.6e-7  // $0.66 per MTok
-	deepseekFlashOffPeakCacheRead   = 7e-9    // $0.007 per MTok (cache hit)
+	deepseekFlashOffPeakInputPrice  = 1.5e-7  // $0.15 per MTok (cache miss)
+	deepseekFlashOffPeakOutputPrice = 6.0e-7  // $0.60 per MTok
+	deepseekFlashOffPeakCacheRead   = 3e-9    // $0.003 per MTok (cache hit)
 	deepseekProOffPeakInputPrice    = 6.6e-7  // $0.66 per MTok (cache miss)
 	deepseekProOffPeakOutputPrice   = 1.98e-6 // $1.98 per MTok
 	deepseekProOffPeakCacheRead     = 2.2e-8  // $0.022 per MTok (cache hit)
@@ -302,6 +309,27 @@ func deepseekPeakMultiplierAt(now time.Time) float64 {
 	return 1.0
 }
 
+// deepseekProRoutesToFlashAt：官方公告自北京时间 2026-09-14 12:00（04:00 UTC）起，
+// 所有 deepseek-v4-pro 请求被上游路由到 V4.1-Flash 并按 Flash 价计费（直至未来
+// V4.1 Pro 上线）。Source: https://api-docs.deepseek.com/news/news260910
+var deepseekProRoutesToFlashAt = time.Date(2026, 9, 14, 4, 0, 0, 0, time.UTC)
+
+// deepseekProBilledAsFlash 报告指定计费时点 deepseek-v4-pro 是否已按 Flash 价
+// 计费：计费时点到达或晚于切换时点返回 true；零值时点回退当前时刻（与峰谷
+// 倍率的取时点方式一致，见 calculateTokenCost）。
+func deepseekProBilledAsFlash(pricingAt time.Time) bool {
+	if pricingAt.IsZero() {
+		pricingAt = timezone.Now()
+	}
+	return !pricingAt.Before(deepseekProRoutesToFlashAt)
+}
+
+// isDeepSeekProModel 判断模型名是否归入 deepseek-v4-pro 档（含版本化名称，
+// 如 deepseek-v4-pro-0813）。
+func isDeepSeekProModel(model string) bool {
+	return strings.Contains(strings.ToLower(strings.TrimSpace(model)), "deepseek-v4-pro")
+}
+
 // BillingService 计费服务
 type BillingService struct {
 	cfg            *config.Config
@@ -557,10 +585,11 @@ func (s *BillingService) initFallbackPricing() {
 
 	// ---- DeepSeek 系列 ----
 	// Source: https://api-docs.deepseek.com/quick_start/pricing
-	// 官方口径（2026-08-23 起生效）：现行模型为 deepseek-v4-flash /
-	// deepseek-v4-pro / deepseek-v4-flash-vision-exp；deepseek-chat /
-	// deepseek-reasoner 已停止服务，其余 deepseek-*（含未知型号）统一按
-	// flash 价兜底（见 getFallbackPricing），避免计费中断。
+	// 官方口径（2026-09-10 公告降价后）：现行模型为 deepseek-flash（=
+	// DeepSeek-V4.1-Flash，旧名 deepseek-v4-flash 兼容路由）/ deepseek-v4-pro /
+	// deepseek-v4-flash-vision-exp；deepseek-chat / deepseek-reasoner 已停止服务，
+	// 其余 deepseek-*（含未知型号）统一按 flash 价兜底（见 getFallbackPricing），
+	// 避免计费中断。
 	// 以下均为官方低谷价；高峰价 = 2× 低谷价（高峰时段 01:00–04:00
 	// 与 06:00–10:00 UTC，仅工作日；北京时间周六/周日全天低谷），见 deepseekPeakMultiplierAt。
 	s.fallbackPrices["deepseek-v4-pro"] = &ModelPricing{
@@ -570,9 +599,9 @@ func (s *BillingService) initFallbackPricing() {
 		SupportsCacheBreakdown: false,
 	}
 	s.fallbackPrices["deepseek-v4-flash"] = &ModelPricing{
-		InputPricePerToken:     deepseekFlashOffPeakInputPrice,  // $0.22 per MTok (cache miss, off-peak)
-		OutputPricePerToken:    deepseekFlashOffPeakOutputPrice, // $0.66 per MTok
-		CacheReadPricePerToken: deepseekFlashOffPeakCacheRead,   // $0.007 per MTok (cache hit)
+		InputPricePerToken:     deepseekFlashOffPeakInputPrice,  // $0.15 per MTok (cache miss, off-peak)
+		OutputPricePerToken:    deepseekFlashOffPeakOutputPrice, // $0.60 per MTok
+		CacheReadPricePerToken: deepseekFlashOffPeakCacheRead,   // $0.003 per MTok (cache hit)
 		SupportsCacheBreakdown: false,
 	}
 	s.fallbackPrices["deepseek-v4-flash-vision-exp"] = &ModelPricing{
@@ -1167,6 +1196,14 @@ func (s *BillingService) HasIdentifiedTokenPricing(model string) bool {
 
 // GetModelPricing 获取模型价格配置
 func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
+	// 无显式计费时点，DeepSeek pro→Flash 切换按当前时刻判定。
+	return s.getModelPricingAt(model, timezone.Now())
+}
+
+// getModelPricingAt 是 GetModelPricing 的带计费时点内部变体：pricingAt 显式
+// 驱动 DeepSeek pro→Flash 切换判定（切换点前 Pro 价、之后 Flash 价），使
+// 展示/估算路径可与历史补账同刻复算，测试也能用固定时点钉住断言。
+func (s *BillingService) getModelPricingAt(model string, pricingAt time.Time) (*ModelPricing, error) {
 	// 标准化模型名称（转小写）
 	model = strings.ToLower(model)
 
@@ -1188,7 +1225,7 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 			price5m := litellmPricing.CacheCreationInputTokenCost
 			price1h := litellmPricing.CacheCreationInputTokenCostAbove1hr
 			enableBreakdown := price1h > 0 && price1h > price5m
-			return s.applyModelSpecificPricingPolicy(model, &ModelPricing{
+			return s.applyModelSpecificPricingPolicyEx(model, &ModelPricing{
 				InputPricePerToken:                 litellmPricing.InputCostPerToken,
 				InputPricePerTokenPriority:         litellmPricing.InputCostPerTokenPriority,
 				OutputPricePerToken:                litellmPricing.OutputCostPerToken,
@@ -1207,7 +1244,7 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 				LongContextOutputMultiplier:   litellmPricing.LongContextOutputCostMultiplier,
 				ImageInputPricePerToken:       litellmPricing.InputCostPerImageToken,
 				ImageOutputPricePerToken:      litellmPricing.OutputCostPerImageToken,
-			}), nil
+			}, true, pricingAt), nil
 		}
 	}
 
@@ -1219,7 +1256,7 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 		if _, seen := s.fallbackWarnSeen.LoadOrStore(model, struct{}{}); !seen {
 			log.Printf("[Billing] Using fallback pricing for model: %s", model)
 		}
-		return s.applyModelSpecificPricingPolicy(model, fallback), nil
+		return s.applyModelSpecificPricingPolicyEx(model, fallback, true, pricingAt), nil
 	}
 
 	return nil, fmt.Errorf("%w for model: %s", ErrModelPricingUnavailable, model)
@@ -1398,20 +1435,22 @@ func (s *BillingService) calculateTokenCost(resolved *ResolvedPricing, input Cos
 		return nil, fmt.Errorf("no pricing available for model: %s: %w", input.Model, ErrModelPricingUnavailable)
 	}
 
+	// 计费时点：优先请求级 PricingAt（历史补账与 DeepSeek pro→Flash 切换判定
+	// 同源），零值回退当前时刻。
+	pricingAt := input.PricingAt
+	if pricingAt.IsZero() {
+		pricingAt = timezone.Now()
+	}
+
 	// 默认价卡（Source=LiteLLM）应用 DeepSeek 官方价强制覆盖（幂等，GetModelPricing
 	// 内部已强制过）；分组/渠道自定义定价保留运营者配置，不强制覆盖官方价。
-	pricing = s.applyModelSpecificPricingPolicyEx(input.Model, pricing, resolved.Source == PricingSourceLiteLLM)
+	pricing = s.applyModelSpecificPricingPolicyEx(input.Model, pricing, resolved.Source == PricingSourceLiteLLM, pricingAt)
 
 	// DeepSeek 模型默认价卡按官方峰谷口径调整：高峰时段（01:00–04:00 与
 	// 06:00–10:00 UTC，仅工作日；北京时间周末全天低谷）按 2× 低谷价计费。
 	// 仅作用于默认价卡（Source=LiteLLM，无分组/渠道自定义定价）——分组/渠道
-	// 自定义定价保持运营者语义，不叠加。PricingAt 为零值时回退当前时刻。
-	// 先克隆再乘，避免污染共享 fallbackPrices 指针。
+	// 自定义定价保持运营者语义，不叠加。先克隆再乘，避免污染共享 fallbackPrices 指针。
 	if resolved.Source == PricingSourceLiteLLM && isDeepSeekModel(input.Model) {
-		pricingAt := input.PricingAt
-		if pricingAt.IsZero() {
-			pricingAt = timezone.Now()
-		}
 		if mult := deepseekPeakMultiplierAt(pricingAt); mult > 1 {
 			cloned := *pricing
 			cloned.InputPricePerToken *= mult
@@ -1670,36 +1709,43 @@ func (s *BillingService) calculateCostInternalWithPolicy(
 // 强制覆盖；GPT-5.6 缺 cache_write 价时按官方规则补 1.25 倍输入价；Fast/priority
 // 档按业务倍率改写（本地/远程目录的 priority 价可能沿用官方旧口径）。长上下文
 // 阶梯不在此处补齐：一律由目录数据（above_XXXk 折算或显式 long_context_* 字段）
-// 驱动。默认强制 DeepSeek 官方价——该路径仅被默认价卡（GetModelPricing 内部）
-// 调用；分组/渠道自定义定价路径用带参数的 applyModelSpecificPricingPolicyEx
-// 关闭强制，保留运营者配置。
+// 驱动。强制 DeepSeek 官方价且无显式计费时点（pro→Flash 切换按当前时刻判定），
+// 供无既有时点的策略修正场景与测试使用；计费/展示主路径分别经
+// calculateTokenCost 与 getModelPricingAt 显式传时点，分组/渠道自定义定价
+// 用 applyModelSpecificPricingPolicyEx 关闭强制，保留运营者配置。
 func (s *BillingService) applyModelSpecificPricingPolicy(model string, pricing *ModelPricing) *ModelPricing {
-	return s.applyModelSpecificPricingPolicyEx(model, pricing, true)
+	return s.applyModelSpecificPricingPolicyEx(model, pricing, true, time.Time{})
 }
 
 // applyModelSpecificPricingPolicyEx 与 applyModelSpecificPricingPolicy 相同，
-// 但由调用方控制是否强制 DeepSeek 官方价（forceDeepSeekRates）。
+// 但由调用方控制是否强制 DeepSeek 官方价（forceDeepSeekRates），并显式传入
+// 计费时点 pricingAt（零值表示按当前时刻判定）。
 // calculateTokenCost 对分组/渠道自定义定价（Source 非 LiteLLM）传 false：
 // 强制覆盖会把运营者配置的售价盖回官方价，违反自定义定价语义。
-func (s *BillingService) applyModelSpecificPricingPolicyEx(model string, pricing *ModelPricing, forceDeepSeekRates bool) *ModelPricing {
+func (s *BillingService) applyModelSpecificPricingPolicyEx(model string, pricing *ModelPricing, forceDeepSeekRates bool, pricingAt time.Time) *ModelPricing {
 	if pricing == nil {
 		return nil
 	}
 	// DeepSeek 模型：无论 JSON/远端价格表给什么价，一律强制官方低谷价
-	// （2026-08-23 起生效）。这是覆盖远端旧价的关键——远端仓库不可改，生产会先
-	// 拉到旧价，必须在此兜底修正；克隆后再覆盖，避免污染共享 fallbackPrices 指针。
+	// （Flash 三档为 2026-09-10 官方降价后口径）。这是覆盖远端旧价的关键——远端
+	// 仓库不可改，生产会先拉到旧价，必须在此兜底修正；克隆后再覆盖，避免污染
+	// 共享 fallbackPrices 指针。
 	// 档位判定：含 "deepseek-v4-pro" 的版本化名称（如 deepseek-v4-pro-0813）归 pro 档，
 	// 其余 deepseek-*（含已停服的 chat/reasoner 与未知型号）统一归 flash 档。
+	// 2026-09-14 04:00 UTC 起上游把 pro 请求路由到 V4.1-Flash，pro 档改按
+	// Flash 三档价计费；历史时点（早于切换时刻）仍按 Pro 价。
 	// 高峰时段倍率不在本函数处理，由 calculateTokenCost 按 deepseekPeakMultiplierAt
 	// 对默认价卡另行叠加（分组/渠道自定义定价不叠加）。
 	if forceDeepSeekRates && isDeepSeekModel(model) {
 		cloned := *pricing
-		if strings.Contains(strings.ToLower(strings.TrimSpace(model)), "deepseek-v4-pro") {
+		if isDeepSeekProModel(model) && !deepseekProBilledAsFlash(pricingAt) {
 			cloned.InputPricePerToken = deepseekProOffPeakInputPrice
 			cloned.OutputPricePerToken = deepseekProOffPeakOutputPrice
 			cloned.CacheReadPricePerToken = deepseekProOffPeakCacheRead
 		} else {
-			// deepseek-v4-flash / deepseek-v4-flash-vision-exp 与其余 deepseek-* 共用 flash 价。
+			// deepseek-flash（= V4.1-Flash）、deepseek-v4-flash /
+			// deepseek-v4-flash-vision-exp 与其余 deepseek-* 共用 flash 价；
+			// 切换时点之后的 pro 请求同样按 flash 价计费。
 			cloned.InputPricePerToken = deepseekFlashOffPeakInputPrice
 			cloned.OutputPricePerToken = deepseekFlashOffPeakOutputPrice
 			cloned.CacheReadPricePerToken = deepseekFlashOffPeakCacheRead
diff --git a/backend/internal/service/billing_service_test.go b/backend/internal/service/billing_service_test.go
index 811419e861..0a68e07ad7 100644
--- a/backend/internal/service/billing_service_test.go
+++ b/backend/internal/service/billing_service_test.go
@@ -514,38 +514,46 @@ func TestGetFallbackPricing_FamilyMatching(t *testing.T) {
 		{
 			name:              "deepseek v4 flash",
 			model:             "deepseek-v4-flash",
-			expectedInput:     2.2e-7,
-			expectedOutput:    floatPtr(6.6e-7),
-			expectedCacheRead: floatPtr(7e-9),
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
+		},
+		{
+			// deepseek-flash（= V4.1-Flash 新名）经前缀兜底同样命中 flash 价卡。
+			name:              "deepseek flash v41 name maps to flash",
+			model:             "deepseek-flash",
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
 		},
 		{
 			name:              "deepseek v4 flash vision exp",
 			model:             "deepseek-v4-flash-vision-exp",
-			expectedInput:     2.2e-7,
-			expectedOutput:    floatPtr(6.6e-7),
-			expectedCacheRead: floatPtr(7e-9),
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
 		},
 		{
 			// deepseek-chat / deepseek-reasoner 已停止服务，统一按 flash 价兜底。
 			name:              "deepseek chat discontinued maps to flash",
 			model:             "deepseek-chat",
-			expectedInput:     2.2e-7,
-			expectedOutput:    floatPtr(6.6e-7),
-			expectedCacheRead: floatPtr(7e-9),
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
 		},
 		{
 			name:              "deepseek reasoner discontinued maps to flash",
 			model:             "deepseek-reasoner",
-			expectedInput:     2.2e-7,
-			expectedOutput:    floatPtr(6.6e-7),
-			expectedCacheRead: floatPtr(7e-9),
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
 		},
 		{
 			name:              "unknown deepseek maps to flash",
 			model:             "deepseek-foo",
-			expectedInput:     2.2e-7,
-			expectedOutput:    floatPtr(6.6e-7),
-			expectedCacheRead: floatPtr(7e-9),
+			expectedInput:     1.5e-7,
+			expectedOutput:    floatPtr(6e-7),
+			expectedCacheRead: floatPtr(3e-9),
 		},
 
 		// ---- 智谱 GLM（z.ai USD 口径）----
diff --git a/backend/internal/service/deepseek_pricing_test.go b/backend/internal/service/deepseek_pricing_test.go
index 3b813929a4..7862d82e0e 100644
--- a/backend/internal/service/deepseek_pricing_test.go
+++ b/backend/internal/service/deepseek_pricing_test.go
@@ -58,7 +58,7 @@ func TestDeepseekPeakMultiplierAt(t *testing.T) {
 
 func TestIsDeepSeekModel(t *testing.T) {
 	deepseek := []string{
-		"deepseek-v4-flash", "deepseek-v4-pro", "deepseek-v4-flash-vision-exp",
+		"deepseek-flash", "deepseek-v4-flash", "deepseek-v4-pro", "deepseek-v4-flash-vision-exp",
 		"deepseek-chat", "deepseek-reasoner", "deepseek-v3-2-251201",
 		"deepseek-coder", "deepseek-foo", "deepseek-v4-pro-0813",
 		"DEEPSEEK-V4-PRO", " deepseek-v4-flash ",
@@ -85,8 +85,8 @@ func TestCalculateCostUnified_DeepseekDefaultCardPeakMultiplier(t *testing.T) {
 	resolver := NewModelPricingResolver(nil, bs)
 
 	tokens := UsageTokens{InputTokens: 1000, OutputTokens: 500, CacheReadTokens: 1000}
-	// 低谷成本：1000*2.2e-7 + 500*6.6e-7 + 1000*7e-9 = 5.57e-4
-	offPeakTotal := 1000*2.2e-7 + 500*6.6e-7 + 1000*7e-9
+	// 低谷成本（2026-09-10 官方新价）：1000*1.5e-7 + 500*6e-7 + 1000*3e-9 = 4.53e-4
+	offPeakTotal := 1000*1.5e-7 + 500*6e-7 + 1000*3e-9
 
 	offPeak, err := bs.CalculateCostUnified(CostInput{
 		Ctx: context.Background(), Model: "deepseek-v4-flash", Tokens: tokens,
@@ -134,7 +134,7 @@ func TestCalculateCostUnified_DeepseekVersionedNamePeakMultiplier(t *testing.T)
 	resolver := NewModelPricingResolver(nil, bs)
 
 	tokens := UsageTokens{InputTokens: 1000, OutputTokens: 500, CacheReadTokens: 1000}
-	offPeakTotal := 1000*2.2e-7 + 500*6.6e-7 + 1000*7e-9
+	offPeakTotal := 1000*1.5e-7 + 500*6e-7 + 1000*3e-9
 
 	offPeak, err := bs.CalculateCostUnified(CostInput{
 		Ctx: context.Background(), Model: "deepseek-v4-flash-0731", Tokens: tokens,
@@ -170,8 +170,8 @@ func TestCalculateCostUnified_DeepseekGroupPricingNotScaledByPeak(t *testing.T)
 	require.Equal(t, PricingSourceGroup, resolved.Source)
 
 	tokens := UsageTokens{InputTokens: 1000, OutputTokens: 500, CacheReadTokens: 1000}
-	// 分组自定义价：1000*1e-6 + 500*2e-6 + 1000*7e-9（缓存读沿用官方 flash 价）
-	groupTotal := 1000*1e-6 + 500*2e-6 + 1000*7e-9
+	// 分组自定义价：1000*1e-6 + 500*2e-6 + 1000*3e-9（缓存读沿用官方 flash 价）
+	groupTotal := 1000*1e-6 + 500*2e-6 + 1000*3e-9
 
 	for _, pricingAt := range []time.Time{
 		time.Date(2026, 8, 24, 12, 0, 0, 0, time.UTC), // 低谷
@@ -238,6 +238,7 @@ func TestCalculateCostUnified_DeepseekPricingAtZeroFallsBackToNow(t *testing.T)
 func TestGetModelPricing_DeepseekForcesOfficialRatesOverJSON(t *testing.T) {
 	// JSON 给任意价（模拟远端旧价/占位价），deepseek-* 必须被强制覆盖为官方低谷价。
 	pricingSvc := &PricingService{pricingData: map[string]*LiteLLMModelPricing{
+		"deepseek-flash":               {InputCostPerToken: 1e-6, OutputCostPerToken: 2e-6, CacheReadInputTokenCost: 1e-8},
 		"deepseek-v4-flash":            {InputCostPerToken: 1e-6, OutputCostPerToken: 2e-6, CacheReadInputTokenCost: 1e-8},
 		"deepseek-v4-pro":              {InputCostPerToken: 1e-6, OutputCostPerToken: 2e-6, CacheReadInputTokenCost: 1e-8},
 		"deepseek-v4-flash-vision-exp": {InputCostPerToken: 1e-6, OutputCostPerToken: 2e-6, CacheReadInputTokenCost: 1e-8},
@@ -250,31 +251,64 @@ func TestGetModelPricing_DeepseekForcesOfficialRatesOverJSON(t *testing.T) {
 		model                    string
 		input, output, cacheRead float64
 	}{
-		{"deepseek-v4-flash", 2.2e-7, 6.6e-7, 7e-9},
-		{"deepseek-v4-flash-vision-exp", 2.2e-7, 6.6e-7, 7e-9},
-		{"deepseek-v4-pro", 6.6e-7, 1.98e-6, 2.2e-8},
+		// 2026-09-10 官方降价后：deepseek-flash（V4.1-Flash 新名）与旧名
+		// deepseek-v4-flash 同按 Flash 新价。
+		{"deepseek-flash", 1.5e-7, 6e-7, 3e-9},
+		{"deepseek-v4-flash", 1.5e-7, 6e-7, 3e-9},
+		{"deepseek-v4-flash-vision-exp", 1.5e-7, 6e-7, 3e-9},
 		// 已停服的 chat/reasoner：即使 JSON 有旧条目也按 flash 价兜底。
-		{"deepseek-chat", 2.2e-7, 6.6e-7, 7e-9},
-		{"deepseek-reasoner", 2.2e-7, 6.6e-7, 7e-9},
+		{"deepseek-chat", 1.5e-7, 6e-7, 3e-9},
+		{"deepseek-reasoner", 1.5e-7, 6e-7, 3e-9},
 	}
 	for _, tt := range tests {
 		t.Run(tt.model, func(t *testing.T) {
+			// flash 档三档价与 pro→Flash 切换无关，GetModelPricing 断言不随时间翻转。
 			pricing, err := bs.GetModelPricing(tt.model)
 			require.NoError(t, err)
 			require.InDelta(t, tt.input, pricing.InputPricePerToken, 1e-15)
 			require.InDelta(t, tt.output, pricing.OutputPricePerToken, 1e-15)
 			require.InDelta(t, tt.cacheRead, pricing.CacheReadPricePerToken, 1e-15)
+			// 固定时点（切换点之后的 2026-10-01）复核：仍走 Flash 新价。
+			atPricing, err := bs.getModelPricingAt(tt.model, time.Date(2026, 10, 1, 0, 0, 0, 0, time.UTC))
+			require.NoError(t, err)
+			require.InDelta(t, tt.input, atPricing.InputPricePerToken, 1e-15)
+			require.InDelta(t, tt.output, atPricing.OutputPricePerToken, 1e-15)
+			require.InDelta(t, tt.cacheRead, atPricing.CacheReadPricePerToken, 1e-15)
 			require.True(t, bs.HasIdentifiedTokenPricing(tt.model))
 		})
 	}
 
+	// pro 档（含版本化名称）：断言经固定时点的 getModelPricingAt，不依赖墙上时钟。
+	// 2026-08-01 早于切换点 2026-09-14 04:00 UTC → Pro 价。
+	for _, model := range []string{"deepseek-v4-pro", "deepseek-v4-pro-0813"} {
+		t.Run(model+"/before-cutoff", func(t *testing.T) {
+			pricing, err := bs.getModelPricingAt(model, time.Date(2026, 8, 1, 0, 0, 0, 0, time.UTC))
+			require.NoError(t, err)
+			require.InDelta(t, 6.6e-7, pricing.InputPricePerToken, 1e-15)
+			require.InDelta(t, 1.98e-6, pricing.OutputPricePerToken, 1e-15)
+			require.InDelta(t, 2.2e-8, pricing.CacheReadPricePerToken, 1e-15)
+		})
+	}
+
+	// 半开边界钉死：2026-09-14 03:59:59 仍 Pro 价，04:00:00 整起 Flash 新价。
+	proBefore, err := bs.getModelPricingAt("deepseek-v4-pro", time.Date(2026, 9, 14, 3, 59, 59, 0, time.UTC))
+	require.NoError(t, err)
+	require.InDelta(t, 6.6e-7, proBefore.InputPricePerToken, 1e-15)
+	require.InDelta(t, 1.98e-6, proBefore.OutputPricePerToken, 1e-15)
+	require.InDelta(t, 2.2e-8, proBefore.CacheReadPricePerToken, 1e-15)
+	proAtCutoff, err := bs.getModelPricingAt("deepseek-v4-pro", time.Date(2026, 9, 14, 4, 0, 0, 0, time.UTC))
+	require.NoError(t, err)
+	require.InDelta(t, 1.5e-7, proAtCutoff.InputPricePerToken, 1e-15)
+	require.InDelta(t, 6e-7, proAtCutoff.OutputPricePerToken, 1e-15)
+	require.InDelta(t, 3e-9, proAtCutoff.CacheReadPricePerToken, 1e-15)
+
 	// 版本化名称（不在 JSON / fallbackPrices 精确表中）：按子串归档计价。
+	// flash-0731 归 flash 档，三档价与切换无关，GetModelPricing 断言稳定。
 	versioned := []struct {
 		model                    string
 		input, output, cacheRead float64
 	}{
-		{"deepseek-v4-pro-0813", 6.6e-7, 1.98e-6, 2.2e-8},
-		{"deepseek-v4-flash-0731", 2.2e-7, 6.6e-7, 7e-9},
+		{"deepseek-v4-flash-0731", 1.5e-7, 6e-7, 3e-9},
 	}
 	for _, tt := range versioned {
 		t.Run(tt.model, func(t *testing.T) {
@@ -289,7 +323,7 @@ func TestGetModelPricing_DeepseekForcesOfficialRatesOverJSON(t *testing.T) {
 
 func TestGetModelPricing_UnknownDeepseekMapsToFlash(t *testing.T) {
 	// JSON 含 $0 占位条目（如旧 deepseek-v3-2-251201）：未知 deepseek-* 不再
-	// fail-closed，统一按 flash 价兜底（2.2e-7/6.6e-7/7e-9），不得按 $0 计费。
+	// fail-closed，统一按 flash 价兜底（1.5e-7/6e-7/3e-9），不得按 $0 计费。
 	pricingSvc := &PricingService{pricingData: map[string]*LiteLLMModelPricing{
 		"deepseek-v3-2-251201": {InputCostPerToken: 0, OutputCostPerToken: 0},
 	}}
@@ -299,9 +333,9 @@ func TestGetModelPricing_UnknownDeepseekMapsToFlash(t *testing.T) {
 		t.Run(m, func(t *testing.T) {
 			pricing, err := bs.GetModelPricing(m)
 			require.NoError(t, err)
-			require.InDelta(t, 2.2e-7, pricing.InputPricePerToken, 1e-15)
-			require.InDelta(t, 6.6e-7, pricing.OutputPricePerToken, 1e-15)
-			require.InDelta(t, 7e-9, pricing.CacheReadPricePerToken, 1e-15)
+			require.InDelta(t, 1.5e-7, pricing.InputPricePerToken, 1e-15)
+			require.InDelta(t, 6e-7, pricing.OutputPricePerToken, 1e-15)
+			require.InDelta(t, 3e-9, pricing.CacheReadPricePerToken, 1e-15)
 		})
 	}
 }
@@ -329,8 +363,9 @@ func TestDeepseekPricingFileMatchesOfficialRates(t *testing.T) {
 		model                    string
 		input, output, cacheRead float64
 	}{
-		{"deepseek-v4-flash", 2.2e-7, 6.6e-7, 7e-9},
-		{"deepseek-v4-flash-vision-exp", 2.2e-7, 6.6e-7, 7e-9},
+		{"deepseek-flash", 1.5e-7, 6e-7, 3e-9},
+		{"deepseek-v4-flash", 1.5e-7, 6e-7, 3e-9},
+		{"deepseek-v4-flash-vision-exp", 1.5e-7, 6e-7, 3e-9},
 		{"deepseek-v4-pro", 6.6e-7, 1.98e-6, 2.2e-8},
 	}
 	for _, tt := range tests {
@@ -343,3 +378,68 @@ func TestDeepseekPricingFileMatchesOfficialRates(t *testing.T) {
 		})
 	}
 }
+
+// ---------------------------------------------------------------------------
+// 2026-09-10 官方降价：deepseek-flash（V4.1-Flash 新名）与旧名同价；
+// 2026-09-14 04:00 UTC 起 deepseek-v4-pro 按上游路由改按 Flash 价计费
+// ---------------------------------------------------------------------------
+
+func TestCalculateCostUnified_DeepseekFlashAndLegacyFlashShareNewRates(t *testing.T) {
+	bs := newTestBillingService()
+	resolver := NewModelPricingResolver(nil, bs)
+
+	tokens := UsageTokens{InputTokens: 1000, OutputTokens: 500, CacheReadTokens: 1000}
+	// 2026-09-10 官方新低谷价：1000*1.5e-7 + 500*6e-7 + 1000*3e-9 = 4.53e-4
+	offPeakTotal := 1000*1.5e-7 + 500*6e-7 + 1000*3e-9
+
+	// deepseek-flash 与 deepseek-v4-flash 都取 Flash 新价。
+	// 时点取切换日 2026-09-14（周一）12:00 UTC 低谷，峰谷倍率不影响断言。
+	for _, model := range []string{"deepseek-flash", "deepseek-v4-flash"} {
+		cost, err := bs.CalculateCostUnified(CostInput{
+			Ctx: context.Background(), Model: model, Tokens: tokens,
+			RateMultiplier: 1.0, Resolver: resolver,
+			PricingAt: time.Date(2026, 9, 14, 12, 0, 0, 0, time.UTC),
+		})
+		require.NoError(t, err)
+		require.InDelta(t, offPeakTotal, cost.TotalCost, 1e-10, "model %s must use new flash rates", model)
+	}
+}
+
+func TestCalculateCostUnified_DeepseekProRoutesToFlashAtCutoff(t *testing.T) {
+	bs := newTestBillingService()
+	resolver := NewModelPricingResolver(nil, bs)
+
+	tokens := UsageTokens{InputTokens: 1000, OutputTokens: 500, CacheReadTokens: 1000}
+	proTotal := 1000*6.6e-7 + 500*1.98e-6 + 1000*2.2e-8 // 切换前 Pro 价
+	flashTotal := 1000*1.5e-7 + 500*6e-7 + 1000*3e-9    // 切换后 Flash 价
+
+	// 切换时点之前（2026-09-13 周日，北京周末全天低谷）：仍按 Pro 价。
+	before, err := bs.CalculateCostUnified(CostInput{
+		Ctx: context.Background(), Model: "deepseek-v4-pro", Tokens: tokens,
+		RateMultiplier: 1.0, Resolver: resolver,
+		PricingAt: time.Date(2026, 9, 13, 12, 0, 0, 0, time.UTC),
+	})
+	require.NoError(t, err)
+	require.InDelta(t, proTotal, before.TotalCost, 1e-10,
+		"deepseek-v4-pro must use Pro rates before the 2026-09-14 04:00 UTC cutoff")
+
+	// 到达切换时点（2026-09-14 04:00 UTC 整，周一低谷窗口边界外）：按 Flash 价。
+	after, err := bs.CalculateCostUnified(CostInput{
+		Ctx: context.Background(), Model: "deepseek-v4-pro", Tokens: tokens,
+		RateMultiplier: 1.0, Resolver: resolver,
+		PricingAt: time.Date(2026, 9, 14, 4, 0, 0, 0, time.UTC),
+	})
+	require.NoError(t, err)
+	require.InDelta(t, flashTotal, after.TotalCost, 1e-10,
+		"deepseek-v4-pro must use Flash rates at/after the 2026-09-14 04:00 UTC cutoff")
+
+	// 版本化名称同口径：切换后 deepseek-v4-pro-0813 也按 Flash 价。
+	versioned, err := bs.CalculateCostUnified(CostInput{
+		Ctx: context.Background(), Model: "deepseek-v4-pro-0813", Tokens: tokens,
+		RateMultiplier: 1.0, Resolver: resolver,
+		PricingAt: time.Date(2026, 9, 14, 4, 0, 0, 0, time.UTC),
+	})
+	require.NoError(t, err)
+	require.InDelta(t, flashTotal, versioned.TotalCost, 1e-10,
+		"versioned pro names must also route to Flash rates after the cutoff")
+}
diff --git a/backend/internal/service/gateway_record_usage_test.go b/backend/internal/service/gateway_record_usage_test.go
index a0d1c0e180..9dfa556298 100644
--- a/backend/internal/service/gateway_record_usage_test.go
+++ b/backend/internal/service/gateway_record_usage_test.go
@@ -465,7 +465,7 @@ func TestGatewayServiceRecordUsage_DeepSeekAccountStatsUsesRequestPricingAtAndUp
 		name        string
 		offPeakCost float64
 	}{
-		{"deepseek-v4-flash", 1000*2.2e-7 + 500*6.6e-7 + 1000*7e-9},
+		{"deepseek-v4-flash", 1000*1.5e-7 + 500*6e-7 + 1000*3e-9},
 		{"deepseek-v4-pro", 1000*6.6e-7 + 500*1.98e-6 + 1000*2.2e-8},
 	} {
 		for _, slot := range []struct {
diff --git a/backend/internal/service/openai_gateway_record_usage_test.go b/backend/internal/service/openai_gateway_record_usage_test.go
index e5dca2d7b7..85b268fbc9 100644
--- a/backend/internal/service/openai_gateway_record_usage_test.go
+++ b/backend/internal/service/openai_gateway_record_usage_test.go
@@ -575,7 +575,7 @@ func TestOpenAIGatewayServiceRecordUsage_DeepSeekAccountStatsUsesRequestPricingA
 		name        string
 		offPeakCost float64
 	}{
-		{"deepseek-v4-flash", 1000*2.2e-7 + 500*6.6e-7 + 1000*7e-9},
+		{"deepseek-v4-flash", 1000*1.5e-7 + 500*6e-7 + 1000*3e-9},
 		{"deepseek-v4-pro", 1000*6.6e-7 + 500*1.98e-6 + 1000*2.2e-8},
 	} {
 		for _, slot := range []struct {
diff --git a/backend/internal/service/openai_gateway_request_body.go b/backend/internal/service/openai_gateway_request_body.go
index af4825762b..c476c9726d 100644
--- a/backend/internal/service/openai_gateway_request_body.go
+++ b/backend/internal/service/openai_gateway_request_body.go
@@ -2180,7 +2180,8 @@ func supportsOpenAIReasoningEffortMax(model string) bool {
 	normalized := strings.ToLower(lastOpenAIModelSegment(model))
 	normalized = strings.ReplaceAll(normalized, "_", "-")
 	switch {
-	case strings.HasPrefix(normalized, "deepseek-v4"):
+	case strings.HasPrefix(normalized, "deepseek-v4"), strings.HasPrefix(normalized, "deepseek-flash"):
+		// deepseek-flash（= DeepSeek-V4.1-Flash）与 v4 系同为 low/high/max 档位。
 		return true
 	case strings.HasPrefix(normalized, "glm-"):
 		return true
diff --git a/backend/internal/service/openai_gpt56_max_test.go b/backend/internal/service/openai_gpt56_max_test.go
index 0b0c8e6fac..fbac20d377 100644
--- a/backend/internal/service/openai_gpt56_max_test.go
+++ b/backend/internal/service/openai_gpt56_max_test.go
@@ -69,6 +69,7 @@ func TestNormalizeOpenAIReasoningEffortForMaxCapableModels(t *testing.T) {
 		{name: "Terra 保留 max", raw: "max", model: "openai/gpt-5.6-terra", want: "max"},
 		{name: "Luna 后缀保留 max", raw: "max", model: "gpt-5.6-luna-2026-07-09", want: "max"},
 		{name: "DeepSeek V4 保留 max", raw: "max", model: "deepseek-v4-pro", want: "max"},
+		{name: "DeepSeek Flash 保留 max", raw: "max", model: "deepseek-flash", want: "max"},
 		{name: "旧 GPT 模型沿用 xhigh", raw: "max", model: "gpt-5.5", want: "xhigh"},
 	}
 
diff --git a/backend/resources/model-pricing/model_prices_and_context_window.json b/backend/resources/model-pricing/model_prices_and_context_window.json
index f971e1c566..579d18db86 100644
--- a/backend/resources/model-pricing/model_prices_and_context_window.json
+++ b/backend/resources/model-pricing/model_prices_and_context_window.json
@@ -745,15 +745,38 @@
     "supports_web_search": true,
     "supports_xhigh_reasoning_effort": true
   },
-  "deepseek-v4-flash": {
-    "cache_read_input_token_cost": 7e-09,
-    "input_cost_per_token": 2.2e-07,
+  "deepseek-flash": {
+    "cache_read_input_token_cost": 3e-09,
+    "input_cost_per_token": 1.5e-07,
     "litellm_provider": "deepseek",
     "max_input_tokens": 1048576,
     "max_output_tokens": 393216,
     "max_tokens": 393216,
     "mode": "chat",
-    "output_cost_per_token": 6.6e-07,
+    "output_cost_per_token": 6e-07,
+    "source": "https://api-docs.deepseek.com/quick_start/pricing",
+    "supported_endpoints": [
+      "/v1/chat/completions"
+    ],
+    "supports_function_calling": true,
+    "supports_native_streaming": true,
+    "supports_parallel_function_calling": true,
+    "supports_prompt_caching": true,
+    "supports_reasoning": true,
+    "supports_response_schema": true,
+    "supports_system_messages": true,
+    "supports_tool_choice": true,
+    "supports_vision": true
+  },
+  "deepseek-v4-flash": {
+    "cache_read_input_token_cost": 3e-09,
+    "input_cost_per_token": 1.5e-07,
+    "litellm_provider": "deepseek",
+    "max_input_tokens": 1048576,
+    "max_output_tokens": 393216,
+    "max_tokens": 393216,
+    "mode": "chat",
+    "output_cost_per_token": 6e-07,
     "source": "https://api-docs.deepseek.com/quick_start/pricing",
     "supported_endpoints": [
       "/v1/chat/completions"
@@ -768,14 +791,14 @@
     "supports_tool_choice": true
   },
   "deepseek-v4-flash-vision-exp": {
-    "cache_read_input_token_cost": 7e-09,
-    "input_cost_per_token": 2.2e-07,
+    "cache_read_input_token_cost": 3e-09,
+    "input_cost_per_token": 1.5e-07,
     "litellm_provider": "deepseek",
     "max_input_tokens": 1048576,
     "max_output_tokens": 393216,
     "max_tokens": 393216,
     "mode": "chat",
-    "output_cost_per_token": 6.6e-07,
+    "output_cost_per_token": 6e-07,
     "source": "https://api-docs.deepseek.com/quick_start/pricing",
     "supported_endpoints": [
       "/v1/chat/completions"
diff --git a/frontend/src/composables/useModelWhitelist.ts b/frontend/src/composables/useModelWhitelist.ts
index d52ede75a4..feed1c37f4 100644
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -116,7 +116,7 @@ const qwenModels = [
 
 // DeepSeek
 const deepseekModels = [
-  'deepseek-v4-pro', 'deepseek-v4-flash', 'deepseek-v4-flash-vision-exp',
+  'deepseek-v4-pro', 'deepseek-v4-flash', 'deepseek-v4-flash-vision-exp', 'deepseek-flash',
   'deepseek-coder',
   'deepseek-v3', 'deepseek-v3-0324',
   'deepseek-r1', 'deepseek-r1-0528',