diff --git a/backend/internal/service/account_usage_service.go b/backend/internal/service/account_usage_service.go
index e89986b5..50deba8b 100644
--- a/backend/internal/service/account_usage_service.go
+++ b/backend/internal/service/account_usage_service.go
@@ -855,6 +855,13 @@ func (s *AccountUsageService) GetAntigravityCredits(ctx context.Context, account
 	return s.getAntigravityUsage(ctx, account)
 }
 
+// InvalidateAntigravityCreditsCache 清除指定账号的 Antigravity 用量缓存，
+// 使下次调用 GetAntigravityCredits 时强制重新拉取。
+// 用于 credits 降级响应重试场景：避免重试命中同一个降级缓存。
+func (s *AccountUsageService) InvalidateAntigravityCreditsCache(accountID int64) {
+	s.cache.antigravityCache.Delete(accountID)
+}
+
 // recalcAntigravityRemainingSeconds 重新计算 Antigravity UsageInfo 中各窗口的 RemainingSeconds
 // 用于从缓存取出时更新倒计时，避免返回过时的剩余秒数
 func recalcAntigravityRemainingSeconds(info *UsageInfo) {
diff --git a/backend/internal/service/antigravity_credits_overages.go b/backend/internal/service/antigravity_credits_overages.go
index 99ec7d08..3e19c563 100644
--- a/backend/internal/service/antigravity_credits_overages.go
+++ b/backend/internal/service/antigravity_credits_overages.go
@@ -4,6 +4,7 @@ import (
 	"context"
 	"encoding/json"
 	"io"
+	"log/slog"
 	"net/http"
 	"strings"
 	"time"
@@ -17,33 +18,116 @@ const (
 	// 与普通模型限流完全同构：通过 SetModelRateLimit / isRateLimitActiveForKey 读写。
 	creditsExhaustedKey      = "AICredits"
 	creditsExhaustedDuration = 5 * time.Hour
+
+	// credits 降级响应重试参数
+	creditsRetryMaxAttempts  = 3
+	creditsRetryBaseInterval = 500 * time.Millisecond
 )
 
+// creditsRetryableErrorCodes 是降级响应中可重试的错误码集合。
+// forbidden 是稳定的封号状态，不属于可恢复的瞬态错误，不重试。
+var creditsRetryableErrorCodes = map[string]bool{
+	errorCodeUnauthenticated: true,
+	errorCodeRateLimited:     true,
+	errorCodeNetworkError:    true,
+}
+
+// isAntigravityDegradedResponse 检查 UsageInfo 是否为可重试的降级响应。
+// 仅检测 3 个瞬态错误码（unauthenticated/rate_limited/network_error），
+// forbidden 是稳定的封号状态，不属于降级。
+func isAntigravityDegradedResponse(info *UsageInfo) bool {
+	if info == nil || info.ErrorCode == "" {
+		return false
+	}
+	return creditsRetryableErrorCodes[info.ErrorCode]
+}
+
 // checkAccountCredits 通过共享的 AccountUsageService 缓存检查账号是否有足够的 AI Credits。
 // 缓存 TTL 不足时会自动从 Google loadCodeAssist API 刷新。
-// 返回 true 表示积分可用。
+// 检测到降级响应时会清除缓存并重试，最终 fail-open（返回 true）。
 func (s *AntigravityGatewayService) checkAccountCredits(
 	ctx context.Context, account *Account,
 ) bool {
 	if account == nil || account.ID == 0 {
 		return false
 	}
-
 	if s.accountUsageService == nil {
 		return true // 无 usage service 时不阻断
 	}
 
 	usageInfo, err := s.accountUsageService.GetAntigravityCredits(ctx, account)
 	if err != nil {
-		logger.LegacyPrintf("service.antigravity_gateway",
-			"check_credits: get_credits_failed account=%d err=%v", account.ID, err)
-		return true // 出错时假设有积分，不阻断
+		slog.Error("check_credits: get_credits_failed",
+			"account_id", account.ID, "error", err)
+		return true // 出错时 fail-open
 	}
 
-	hasCredits := hasEnoughCredits(usageInfo)
+	// 非降级响应：直接检查积分余额
+	if !isAntigravityDegradedResponse(usageInfo) {
+		return s.logCreditsResult(account, usageInfo)
+	}
+
+	// 降级响应：清除缓存后重试
+	return s.retryCreditsOnDegraded(ctx, account, usageInfo)
+}
+
+// retryCreditsOnDegraded 在检测到降级响应后，清除缓存并重试获取 credits。
+// 使用指数退避（500ms → 1s → 2s），最多重试 creditsRetryMaxAttempts 次。
+// 所有重试失败后 fail-open（返回 true），不做熔断。
+func (s *AntigravityGatewayService) retryCreditsOnDegraded(
+	ctx context.Context, account *Account, lastInfo *UsageInfo,
+) bool {
+	for attempt := 1; attempt <= creditsRetryMaxAttempts; attempt++ {
+		delay := creditsRetryBaseInterval << (attempt - 1) // 指数退避：500ms, 1s, 2s
+		slog.Warn("check_credits: degraded response, retrying",
+			"account_id", account.ID,
+			"attempt", attempt,
+			"max_attempts", creditsRetryMaxAttempts,
+			"error_code", lastInfo.ErrorCode,
+			"delay", delay,
+		)
+
+		select {
+		case <-ctx.Done():
+			slog.Warn("check_credits: context cancelled during retry, fail-open",
+				"account_id", account.ID, "attempt", attempt)
+			return true
+		case <-time.After(delay):
+		}
+
+		// 清除缓存，强制下次 GetAntigravityCredits 重新拉取
+		s.accountUsageService.InvalidateAntigravityCreditsCache(account.ID)
+
+		info, err := s.accountUsageService.GetAntigravityCredits(ctx, account)
+		if err != nil {
+			slog.Error("check_credits: retry get_credits_failed",
+				"account_id", account.ID, "attempt", attempt, "error", err)
+			continue
+		}
+
+		// 重试成功（不再是降级响应）：检查积分余额
+		if !isAntigravityDegradedResponse(info) {
+			slog.Info("check_credits: retry succeeded",
+				"account_id", account.ID, "attempt", attempt)
+			return s.logCreditsResult(account, info)
+		}
+		lastInfo = info
+	}
+
+	// 所有重试失败：fail-open，不做熔断
+	slog.Warn("check_credits: all retries exhausted, fail-open",
+		"account_id", account.ID,
+		"last_error_code", lastInfo.ErrorCode,
+	)
+	return true
+}
+
+// logCreditsResult 检查积分并记录不足日志，返回是否有积分。
+func (s *AntigravityGatewayService) logCreditsResult(account *Account, info *UsageInfo) bool {
+	hasCredits := hasEnoughCredits(info)
 	if !hasCredits {
-		logger.LegacyPrintf("service.antigravity_gateway",
-			"check_credits: account=%d has_credits=false", account.ID)
+		slog.Warn("check_credits: insufficient credits",
+			"account_id", account.ID)
 	}
 	return hasCredits
 }
diff --git a/backend/internal/service/channel.go b/backend/internal/service/channel.go
index 05f96bc9..1697ed6f 100644
--- a/backend/internal/service/channel.go
+++ b/backend/internal/service/channel.go
@@ -1,6 +1,8 @@
 package service
 
 import (
+	"fmt"
+	"sort"
 	"strings"
 	"time"
 )
@@ -177,6 +179,94 @@ func (c *Channel) Clone() *Channel {
 	return &cp
 }
 
+// ValidateIntervals 校验区间列表的合法性。
+// 规则：MinTokens >= 0；MaxTokens 若非 nil 则 > 0 且 > MinTokens；
+// 所有价格字段 >= 0；区间按 MinTokens 排序后无重叠（(min, max] 语义）；
+// 无界区间（MaxTokens=nil）必须是最后一个。间隙允许（回退默认价格）。
+func ValidateIntervals(intervals []PricingInterval) error {
+	if len(intervals) == 0 {
+		return nil
+	}
+	sorted := make([]PricingInterval, len(intervals))
+	copy(sorted, intervals)
+	sort.Slice(sorted, func(i, j int) bool {
+		return sorted[i].MinTokens < sorted[j].MinTokens
+	})
+
+	for i := range sorted {
+		if err := validateSingleInterval(&sorted[i], i); err != nil {
+			return err
+		}
+	}
+	return validateIntervalOverlap(sorted)
+}
+
+// validateSingleInterval 校验单个区间的字段合法性
+func validateSingleInterval(iv *PricingInterval, idx int) error {
+	if iv.MinTokens < 0 {
+		return fmt.Errorf("interval #%d: min_tokens (%d) must be >= 0", idx+1, iv.MinTokens)
+	}
+	if iv.MaxTokens != nil {
+		if *iv.MaxTokens <= 0 {
+			return fmt.Errorf("interval #%d: max_tokens (%d) must be > 0", idx+1, *iv.MaxTokens)
+		}
+		if *iv.MaxTokens <= iv.MinTokens {
+			return fmt.Errorf("interval #%d: max_tokens (%d) must be > min_tokens (%d)",
+				idx+1, *iv.MaxTokens, iv.MinTokens)
+		}
+	}
+	return validateIntervalPrices(iv, idx)
+}
+
+// validateIntervalPrices 校验区间内所有价格字段 >= 0
+func validateIntervalPrices(iv *PricingInterval, idx int) error {
+	prices := []struct {
+		name string
+		val  *float64
+	}{
+		{"input_price", iv.InputPrice},
+		{"output_price", iv.OutputPrice},
+		{"cache_write_price", iv.CacheWritePrice},
+		{"cache_read_price", iv.CacheReadPrice},
+		{"per_request_price", iv.PerRequestPrice},
+	}
+	for _, p := range prices {
+		if p.val != nil && *p.val < 0 {
+			return fmt.Errorf("interval #%d: %s must be >= 0", idx+1, p.name)
+		}
+	}
+	return nil
+}
+
+// validateIntervalOverlap 校验排序后的区间列表无重叠，且无界区间在最后
+func validateIntervalOverlap(sorted []PricingInterval) error {
+	for i, iv := range sorted {
+		// 无界区间必须是最后一个
+		if iv.MaxTokens == nil && i < len(sorted)-1 {
+			return fmt.Errorf("interval #%d: unbounded interval (max_tokens=null) must be the last one",
+				i+1)
+		}
+		if i == 0 {
+			continue
+		}
+		prev := sorted[i-1]
+		// 检查重叠：前一个区间的上界 > 当前区间的下界则重叠
+		// (min, max] 语义：prev 覆盖 (prev.Min, prev.Max]，cur 覆盖 (cur.Min, cur.Max]
+		if prev.MaxTokens == nil || *prev.MaxTokens > iv.MinTokens {
+			return fmt.Errorf("interval #%d and #%d overlap: prev max=%s > cur min=%d",
+				i, i+1, formatMaxTokensLabel(prev.MaxTokens), iv.MinTokens)
+		}
+	}
+	return nil
+}
+
+func formatMaxTokensLabel(max *int) string {
+	if max == nil {
+		return "∞"
+	}
+	return fmt.Sprintf("%d", *max)
+}
+
 // ChannelUsageFields 渠道相关的使用记录字段（嵌入到各平台的 RecordUsageInput 中）
 type ChannelUsageFields struct {
 	ChannelID          int64  // 渠道 ID（0 = 无渠道）
diff --git a/backend/internal/service/channel_service.go b/backend/internal/service/channel_service.go
index 25a8d39b..cbab9bfe 100644
--- a/backend/internal/service/channel_service.go
+++ b/backend/internal/service/channel_service.go
@@ -198,13 +198,18 @@ func newEmptyChannelCache() *channelCache {
 
 // expandPricingToCache 将渠道的模型定价展开到缓存（按分组+平台维度）。
 // antigravity 平台同时服务 Claude 和 Gemini 模型，需匹配 anthropic/gemini 的定价条目。
+// 缓存 key 使用定价条目的原始平台（pricing.Platform），而非分组平台，
+// 避免跨平台同名模型（如 anthropic 和 gemini 都有 "model-x"）互相覆盖。
+// 查找时通过 lookupPricingAcrossPlatforms() 依次尝试所有匹配平台。
 func expandPricingToCache(cache *channelCache, ch *Channel, gid int64, platform string) {
 	for j := range ch.ModelPricing {
 		pricing := &ch.ModelPricing[j]
 		if !isPlatformPricingMatch(platform, pricing.Platform) {
 			continue // 跳过非本平台的定价
 		}
-		gpKey := channelGroupPlatformKey{groupID: gid, platform: platform}
+		// 使用定价条目的原始平台作为缓存 key，防止跨平台同名模型冲突
+		pricingPlatform := pricing.Platform
+		gpKey := channelGroupPlatformKey{groupID: gid, platform: pricingPlatform}
 		for _, model := range pricing.Models {
 			if strings.HasSuffix(model, "*") {
 				prefix := strings.ToLower(strings.TrimSuffix(model, "*"))
@@ -213,7 +218,7 @@ func expandPricingToCache(cache *channelCache, ch *Channel, gid int64, platform
 					pricing: pricing,
 				})
 			} else {
-				key := channelModelKey{groupID: gid, platform: platform, model: strings.ToLower(model)}
+				key := channelModelKey{groupID: gid, platform: pricingPlatform, model: strings.ToLower(model)}
 				cache.pricingByGroupModel[key] = pricing
 			}
 		}
@@ -222,13 +227,15 @@ func expandPricingToCache(cache *channelCache, ch *Channel, gid int64, platform
 
 // expandMappingToCache 将渠道的模型映射展开到缓存（按分组+平台维度）。
 // antigravity 平台同时服务 Claude 和 Gemini 模型。
+// 缓存 key 使用映射条目的原始平台（mappingPlatform），避免跨平台同名映射覆盖。
 func expandMappingToCache(cache *channelCache, ch *Channel, gid int64, platform string) {
 	for _, mappingPlatform := range matchingPlatforms(platform) {
 		platformMapping, ok := ch.ModelMapping[mappingPlatform]
 		if !ok {
 			continue
 		}
-		gpKey := channelGroupPlatformKey{groupID: gid, platform: platform}
+		// 使用映射条目的原始平台作为缓存 key，防止跨平台同名映射冲突
+		gpKey := channelGroupPlatformKey{groupID: gid, platform: mappingPlatform}
 		for src, dst := range platformMapping {
 			if strings.HasSuffix(src, "*") {
 				prefix := strings.ToLower(strings.TrimSuffix(src, "*"))
@@ -237,7 +244,7 @@ func expandMappingToCache(cache *channelCache, ch *Channel, gid int64, platform
 					target: dst,
 				})
 			} else {
-				key := channelModelKey{groupID: gid, platform: platform, model: strings.ToLower(src)}
+				key := channelModelKey{groupID: gid, platform: mappingPlatform, model: strings.ToLower(src)}
 				cache.mappingByGroupModel[key] = dst
 			}
 		}
@@ -349,6 +356,43 @@ func (c *channelCache) matchWildcardMapping(groupID int64, platform, modelLower
 	return ""
 }
 
+// lookupPricingAcrossPlatforms 在所有匹配平台中查找模型定价。
+// antigravity 分组的缓存 key 使用定价条目的原始平台，因此查找时需依次尝试
+// matchingPlatforms() 返回的所有平台（antigravity → anthropic → gemini），
+// 返回第一个命中的结果。非 antigravity 平台只尝试自身。
+func lookupPricingAcrossPlatforms(cache *channelCache, groupID int64, groupPlatform, modelLower string) *ChannelModelPricing {
+	for _, p := range matchingPlatforms(groupPlatform) {
+		key := channelModelKey{groupID: groupID, platform: p, model: modelLower}
+		if pricing, ok := cache.pricingByGroupModel[key]; ok {
+			return pricing
+		}
+	}
+	// 精确查找全部失败，依次尝试通配符匹配
+	for _, p := range matchingPlatforms(groupPlatform) {
+		if pricing := cache.matchWildcard(groupID, p, modelLower); pricing != nil {
+			return pricing
+		}
+	}
+	return nil
+}
+
+// lookupMappingAcrossPlatforms 在所有匹配平台中查找模型映射。
+// 逻辑与 lookupPricingAcrossPlatforms 相同：先精确查找，再通配符。
+func lookupMappingAcrossPlatforms(cache *channelCache, groupID int64, groupPlatform, modelLower string) string {
+	for _, p := range matchingPlatforms(groupPlatform) {
+		key := channelModelKey{groupID: groupID, platform: p, model: modelLower}
+		if mapped, ok := cache.mappingByGroupModel[key]; ok {
+			return mapped
+		}
+	}
+	for _, p := range matchingPlatforms(groupPlatform) {
+		if mapped := cache.matchWildcardMapping(groupID, p, modelLower); mapped != "" {
+			return mapped
+		}
+	}
+	return ""
+}
+
 // GetChannelForGroup 获取分组关联的渠道（热路径 O(1)）
 func (s *ChannelService) GetChannelForGroup(ctx context.Context, groupID int64) (*Channel, error) {
 	cache, err := s.loadCache(ctx)
@@ -389,7 +433,9 @@ func (s *ChannelService) lookupGroupChannel(ctx context.Context, groupID int64)
 	}, nil
 }
 
-// GetChannelModelPricing 获取指定分组+模型的渠道定价（热路径 O(1)）
+// GetChannelModelPricing 获取指定分组+模型的渠道定价（热路径 O(1)）。
+// antigravity 分组依次尝试所有匹配平台（antigravity → anthropic → gemini），
+// 确保跨平台同名模型各自独立匹配。
 func (s *ChannelService) GetChannelModelPricing(ctx context.Context, groupID int64, model string) *ChannelModelPricing {
 	lk, err := s.lookupGroupChannel(ctx, groupID)
 	if err != nil {
@@ -401,14 +447,9 @@ func (s *ChannelService) GetChannelModelPricing(ctx context.Context, groupID int
 	}
 
 	modelLower := strings.ToLower(model)
-	key := channelModelKey{groupID: groupID, platform: lk.platform, model: modelLower}
-	pricing, ok := lk.cache.pricingByGroupModel[key]
-	if !ok {
-		// 精确查找失败，尝试通配符匹配
-		pricing = lk.cache.matchWildcard(groupID, lk.platform, modelLower)
-		if pricing == nil {
-			return nil
-		}
+	pricing := lookupPricingAcrossPlatforms(lk.cache, groupID, lk.platform, modelLower)
+	if pricing == nil {
+		return nil
 	}
 
 	cp := pricing.Clone()
@@ -453,7 +494,8 @@ func (s *ChannelService) ResolveChannelMappingAndRestrict(ctx context.Context, g
 	return resolveMapping(lk, *groupID, model), false
 }
 
-// resolveMapping 基于已查找的渠道信息解析模型映射
+// resolveMapping 基于已查找的渠道信息解析模型映射。
+// antigravity 分组依次尝试所有匹配平台，确保跨平台同名映射各自独立。
 func resolveMapping(lk *channelLookup, groupID int64, model string) ChannelMappingResult {
 	result := ChannelMappingResult{
 		MappedModel:        model,
@@ -465,11 +507,7 @@ func resolveMapping(lk *channelLookup, groupID int64, model string) ChannelMappi
 	}
 
 	modelLower := strings.ToLower(model)
-	key := channelModelKey{groupID: groupID, platform: lk.platform, model: modelLower}
-	if mapped, ok := lk.cache.mappingByGroupModel[key]; ok {
-		result.MappedModel = mapped
-		result.Mapped = true
-	} else if mapped := lk.cache.matchWildcardMapping(groupID, lk.platform, modelLower); mapped != "" {
+	if mapped := lookupMappingAcrossPlatforms(lk.cache, groupID, lk.platform, modelLower); mapped != "" {
 		result.MappedModel = mapped
 		result.Mapped = true
 	}
@@ -477,19 +515,15 @@ func resolveMapping(lk *channelLookup, groupID int64, model string) ChannelMappi
 	return result
 }
 
-// checkRestricted 基于已查找的渠道信息检查模型是否被限制
+// checkRestricted 基于已查找的渠道信息检查模型是否被限制。
+// antigravity 分组依次尝试所有匹配平台的定价列表。
 func checkRestricted(lk *channelLookup, groupID int64, model string) bool {
 	if !lk.channel.RestrictModels {
 		return false
 	}
-	// 检查模型是否在定价列表中
 	modelLower := strings.ToLower(model)
-	key := channelModelKey{groupID: groupID, platform: lk.platform, model: modelLower}
-	if _, exists := lk.cache.pricingByGroupModel[key]; exists {
-		return false
-	}
-	// 精确查找失败，尝试通配符匹配
-	if lk.cache.matchWildcard(groupID, lk.platform, modelLower) != nil {
+	// 使用与查找定价相同的跨平台逻辑
+	if lookupPricingAcrossPlatforms(lk.cache, groupID, lk.platform, modelLower) != nil {
 		return false
 	}
 	return true
@@ -550,6 +584,9 @@ func (s *ChannelService) Create(ctx context.Context, input *CreateChannelInput)
 	if err := validateNoConflictingModels(channel.ModelPricing); err != nil {
 		return nil, err
 	}
+	if err := validatePricingIntervals(channel.ModelPricing); err != nil {
+		return nil, err
+	}
 	if err := validateNoConflictingMappings(channel.ModelMapping); err != nil {
 		return nil, err
 	}
@@ -624,6 +661,9 @@ func (s *ChannelService) Update(ctx context.Context, id int64, input *UpdateChan
 	if err := validateNoConflictingModels(channel.ModelPricing); err != nil {
 		return nil, err
 	}
+	if err := validatePricingIntervals(channel.ModelPricing); err != nil {
+		return nil, err
+	}
 	if err := validateNoConflictingMappings(channel.ModelMapping); err != nil {
 		return nil, err
 	}
@@ -756,6 +796,19 @@ func validateNoConflictingMappings(mapping map[string]map[string]string) error {
 	return nil
 }
 
+func validatePricingIntervals(pricingList []ChannelModelPricing) error {
+	for _, pricing := range pricingList {
+		if err := ValidateIntervals(pricing.Intervals); err != nil {
+			return infraerrors.BadRequest(
+				"INVALID_PRICING_INTERVALS",
+				fmt.Sprintf("invalid pricing intervals for platform '%s' models %v: %v",
+					pricing.Platform, pricing.Models, err),
+			)
+		}
+	}
+	return nil
+}
+
 // detectConflicts 在一组 modelEntry 中检测冲突，返回带有 errCode 和 label 的错误
 func detectConflicts(entries []modelEntry, platform, errCode, label string) error {
 	for i := 0; i < len(entries); i++ {
diff --git a/backend/internal/service/channel_service_test.go b/backend/internal/service/channel_service_test.go
index 93cafa67..0232062c 100644
--- a/backend/internal/service/channel_service_test.go
+++ b/backend/internal/service/channel_service_test.go
@@ -1401,6 +1401,32 @@ func TestCreate_DuplicateModel(t *testing.T) {
 	require.Contains(t, err.Error(), "claude-opus-4")
 }
 
+func TestCreate_InvalidPricingIntervals(t *testing.T) {
+	repo := &mockChannelRepository{
+		existsByNameFn: func(_ context.Context, _ string) (bool, error) {
+			return false, nil
+		},
+	}
+	svc := newTestChannelService(repo)
+
+	_, err := svc.Create(context.Background(), &CreateChannelInput{
+		Name: "new-channel",
+		ModelPricing: []ChannelModelPricing{
+			{
+				Platform: "anthropic",
+				Models:   []string{"claude-opus-4"},
+				Intervals: []PricingInterval{
+					{MinTokens: 0, MaxTokens: testPtrInt(2000), InputPrice: testPtrFloat64(1e-6)},
+					{MinTokens: 1000, MaxTokens: testPtrInt(3000), InputPrice: testPtrFloat64(2e-6)},
+				},
+			},
+		},
+	})
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "INVALID_PRICING_INTERVALS")
+	require.Contains(t, err.Error(), "overlap")
+}
+
 func TestCreate_DefaultBillingModelSource(t *testing.T) {
 	var capturedChannel *Channel
 	repo := &mockChannelRepository{
@@ -1592,6 +1618,37 @@ func TestUpdate_DuplicateModel(t *testing.T) {
 	require.Contains(t, err.Error(), "claude-opus-4")
 }
 
+func TestUpdate_InvalidPricingIntervals(t *testing.T) {
+	existing := &Channel{
+		ID:     1,
+		Name:   "original",
+		Status: StatusActive,
+	}
+	repo := &mockChannelRepository{
+		getByIDFn: func(_ context.Context, _ int64) (*Channel, error) {
+			return existing.Clone(), nil
+		},
+	}
+	svc := newTestChannelService(repo)
+
+	invalidPricing := []ChannelModelPricing{
+		{
+			Platform: "anthropic",
+			Models:   []string{"claude-opus-4"},
+			Intervals: []PricingInterval{
+				{MinTokens: 0, MaxTokens: nil, InputPrice: testPtrFloat64(1e-6)},
+				{MinTokens: 2000, MaxTokens: testPtrInt(4000), InputPrice: testPtrFloat64(2e-6)},
+			},
+		},
+	}
+	_, err := svc.Update(context.Background(), 1, &UpdateChannelInput{
+		ModelPricing: &invalidPricing,
+	})
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "INVALID_PRICING_INTERVALS")
+	require.Contains(t, err.Error(), "unbounded")
+}
+
 func TestUpdate_InvalidatesChannelCache(t *testing.T) {
 	existing := &Channel{
 		ID:     1,
@@ -1984,3 +2041,144 @@ func TestResolveChannelMapping_AntigravityCrossPlatform(t *testing.T) {
 	require.Equal(t, "claude-opus-4-6", result.MappedModel)
 	require.Equal(t, int64(1), result.ChannelID)
 }
+
+// ===========================================================================
+// 11. Antigravity cross-platform same-name model — no overwrite
+// ===========================================================================
+
+func TestGetChannelModelPricing_AntigravitySameModelDifferentPlatforms(t *testing.T) {
+	// anthropic 和 gemini 都定义了同名模型 "shared-model"，价格不同。
+	// antigravity 分组应能分别查到各自的定价，而不是后者覆盖前者。
+	ch := Channel{
+		ID:       1,
+		Status:   StatusActive,
+		GroupIDs: []int64{10},
+		ModelPricing: []ChannelModelPricing{
+			{ID: 200, Platform: PlatformAnthropic, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(10e-6)},
+			{ID: 201, Platform: PlatformGemini, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(5e-6)},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAntigravity})
+	svc := newTestChannelService(repo)
+
+	// antigravity 分组查找 "shared-model"：应命中第一个匹配（按 matchingPlatforms 顺序 antigravity→anthropic→gemini）
+	result := svc.GetChannelModelPricing(context.Background(), 10, "shared-model")
+	require.NotNil(t, result, "antigravity group should find pricing for shared-model")
+	// 第一个匹配应该是 anthropic（matchingPlatforms 返回 [antigravity, anthropic, gemini]）
+	require.Equal(t, int64(200), result.ID)
+	require.InDelta(t, 10e-6, *result.InputPrice, 1e-12)
+}
+
+func TestGetChannelModelPricing_AntigravityOnlyGeminiPricing(t *testing.T) {
+	// 只有 gemini 平台定义了模型 "gemini-model"。
+	// antigravity 分组应能查到 gemini 的定价。
+	ch := Channel{
+		ID:       1,
+		Status:   StatusActive,
+		GroupIDs: []int64{10},
+		ModelPricing: []ChannelModelPricing{
+			{ID: 300, Platform: PlatformGemini, Models: []string{"gemini-model"}, InputPrice: testPtrFloat64(2e-6)},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAntigravity})
+	svc := newTestChannelService(repo)
+
+	result := svc.GetChannelModelPricing(context.Background(), 10, "gemini-model")
+	require.NotNil(t, result, "antigravity group should find gemini pricing")
+	require.Equal(t, int64(300), result.ID)
+	require.InDelta(t, 2e-6, *result.InputPrice, 1e-12)
+}
+
+func TestGetChannelModelPricing_AntigravityWildcardCrossPlatformNoOverwrite(t *testing.T) {
+	// anthropic 和 gemini 都有 "shared-*" 通配符定价，价格不同。
+	// antigravity 分组查找 "shared-model" 应命中第一个匹配而非被覆盖。
+	ch := Channel{
+		ID:       1,
+		Status:   StatusActive,
+		GroupIDs: []int64{10},
+		ModelPricing: []ChannelModelPricing{
+			{ID: 400, Platform: PlatformAnthropic, Models: []string{"shared-*"}, InputPrice: testPtrFloat64(10e-6)},
+			{ID: 401, Platform: PlatformGemini, Models: []string{"shared-*"}, InputPrice: testPtrFloat64(5e-6)},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAntigravity})
+	svc := newTestChannelService(repo)
+
+	result := svc.GetChannelModelPricing(context.Background(), 10, "shared-model")
+	require.NotNil(t, result, "antigravity group should find wildcard pricing for shared-model")
+	// 两个通配符都存在，应命中 anthropic 的（matchingPlatforms 顺序）
+	require.Equal(t, int64(400), result.ID)
+	require.InDelta(t, 10e-6, *result.InputPrice, 1e-12)
+}
+
+func TestResolveChannelMapping_AntigravitySameModelDifferentPlatforms(t *testing.T) {
+	// anthropic 和 gemini 都定义了同名模型映射 "alias" → 不同目标。
+	// antigravity 分组应命中 anthropic 的映射（按 matchingPlatforms 顺序）。
+	ch := Channel{
+		ID:       1,
+		Status:   StatusActive,
+		GroupIDs: []int64{10},
+		ModelMapping: map[string]map[string]string{
+			PlatformAnthropic: {"alias": "anthropic-target"},
+			PlatformGemini:    {"alias": "gemini-target"},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAntigravity})
+	svc := newTestChannelService(repo)
+
+	result := svc.ResolveChannelMapping(context.Background(), 10, "alias")
+	require.True(t, result.Mapped)
+	require.Equal(t, "anthropic-target", result.MappedModel)
+}
+
+func TestCheckRestricted_AntigravitySameModelDifferentPlatforms(t *testing.T) {
+	// anthropic 和 gemini 都定义了同名模型 "shared-model"。
+	// antigravity 分组启用了 RestrictModels，"shared-model" 应不被限制。
+	ch := Channel{
+		ID:             1,
+		Status:         StatusActive,
+		RestrictModels: true,
+		GroupIDs:       []int64{10},
+		ModelPricing: []ChannelModelPricing{
+			{ID: 500, Platform: PlatformAnthropic, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(10e-6)},
+			{ID: 501, Platform: PlatformGemini, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(5e-6)},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAntigravity})
+	svc := newTestChannelService(repo)
+
+	restricted := svc.IsModelRestricted(context.Background(), 10, "shared-model")
+	require.False(t, restricted, "shared-model should not be restricted for antigravity")
+
+	// 未定义的模型应被限制
+	restricted = svc.IsModelRestricted(context.Background(), 10, "unknown-model")
+	require.True(t, restricted, "unknown-model should be restricted for antigravity")
+}
+
+func TestGetChannelModelPricing_NonAntigravityUnaffected(t *testing.T) {
+	// 确保非 antigravity 平台的行为不受影响。
+	// anthropic 分组只能看到 anthropic 的定价，看不到 gemini 的。
+	ch := Channel{
+		ID:       1,
+		Status:   StatusActive,
+		GroupIDs: []int64{10, 20},
+		ModelPricing: []ChannelModelPricing{
+			{ID: 600, Platform: PlatformAnthropic, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(10e-6)},
+			{ID: 601, Platform: PlatformGemini, Models: []string{"shared-model"}, InputPrice: testPtrFloat64(5e-6)},
+		},
+	}
+	repo := makeStandardRepo(ch, map[int64]string{10: PlatformAnthropic, 20: PlatformGemini})
+	svc := newTestChannelService(repo)
+
+	// anthropic 分组应该只看到 anthropic 的定价
+	result := svc.GetChannelModelPricing(context.Background(), 10, "shared-model")
+	require.NotNil(t, result)
+	require.Equal(t, int64(600), result.ID)
+	require.InDelta(t, 10e-6, *result.InputPrice, 1e-12)
+
+	// gemini 分组应该只看到 gemini 的定价
+	result = svc.GetChannelModelPricing(context.Background(), 20, "shared-model")
+	require.NotNil(t, result)
+	require.Equal(t, int64(601), result.ID)
+	require.InDelta(t, 5e-6, *result.InputPrice, 1e-12)
+}
diff --git a/backend/internal/service/channel_test.go b/backend/internal/service/channel_test.go
index d01c252b..deac64d6 100644
--- a/backend/internal/service/channel_test.go
+++ b/backend/internal/service/channel_test.go
@@ -307,3 +307,129 @@ func TestChannelClone_EdgeCases(t *testing.T) {
 		require.Equal(t, "gpt-4-turbo", original.ModelMapping["openai"]["gpt-4"])
 	})
 }
+
+// --- ValidateIntervals ---
+
+func TestValidateIntervals_Empty(t *testing.T) {
+	require.NoError(t, ValidateIntervals(nil))
+	require.NoError(t, ValidateIntervals([]PricingInterval{}))
+}
+
+func TestValidateIntervals_ValidIntervals(t *testing.T) {
+	tests := []struct {
+		name      string
+		intervals []PricingInterval
+	}{
+		{
+			name: "single bounded interval",
+			intervals: []PricingInterval{
+				{MinTokens: 0, MaxTokens: testPtrInt(128000), InputPrice: testPtrFloat64(1e-6)},
+			},
+		},
+		{
+			name: "two intervals with gap",
+			intervals: []PricingInterval{
+				{MinTokens: 0, MaxTokens: testPtrInt(100000), InputPrice: testPtrFloat64(1e-6)},
+				{MinTokens: 128000, MaxTokens: nil, InputPrice: testPtrFloat64(2e-6)},
+			},
+		},
+		{
+			name: "two contiguous intervals",
+			intervals: []PricingInterval{
+				{MinTokens: 0, MaxTokens: testPtrInt(128000), InputPrice: testPtrFloat64(1e-6)},
+				{MinTokens: 128000, MaxTokens: nil, InputPrice: testPtrFloat64(2e-6)},
+			},
+		},
+		{
+			name: "unsorted input (auto-sorted by validator)",
+			intervals: []PricingInterval{
+				{MinTokens: 128000, MaxTokens: nil, InputPrice: testPtrFloat64(2e-6)},
+				{MinTokens: 0, MaxTokens: testPtrInt(128000), InputPrice: testPtrFloat64(1e-6)},
+			},
+		},
+		{
+			name: "single unbounded interval",
+			intervals: []PricingInterval{
+				{MinTokens: 0, MaxTokens: nil, InputPrice: testPtrFloat64(1e-6)},
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			require.NoError(t, ValidateIntervals(tt.intervals))
+		})
+	}
+}
+
+func TestValidateIntervals_NegativeMinTokens(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: -1, MaxTokens: testPtrInt(100), InputPrice: testPtrFloat64(1e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "min_tokens")
+	require.Contains(t, err.Error(), ">= 0")
+}
+
+func TestValidateIntervals_MaxTokensZero(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: 0, MaxTokens: testPtrInt(0), InputPrice: testPtrFloat64(1e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "max_tokens")
+	require.Contains(t, err.Error(), "> 0")
+}
+
+func TestValidateIntervals_MaxLessThanMin(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: 100, MaxTokens: testPtrInt(50), InputPrice: testPtrFloat64(1e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "max_tokens")
+	require.Contains(t, err.Error(), "> min_tokens")
+}
+
+func TestValidateIntervals_MaxEqualsMin(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: 100, MaxTokens: testPtrInt(100), InputPrice: testPtrFloat64(1e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "max_tokens")
+	require.Contains(t, err.Error(), "> min_tokens")
+}
+
+func TestValidateIntervals_NegativePrice(t *testing.T) {
+	negPrice := -0.01
+	intervals := []PricingInterval{
+		{MinTokens: 0, MaxTokens: testPtrInt(100), InputPrice: &negPrice},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "input_price")
+	require.Contains(t, err.Error(), ">= 0")
+}
+
+func TestValidateIntervals_OverlappingIntervals(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: 0, MaxTokens: testPtrInt(200), InputPrice: testPtrFloat64(1e-6)},
+		{MinTokens: 100, MaxTokens: testPtrInt(300), InputPrice: testPtrFloat64(2e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "overlap")
+}
+
+func TestValidateIntervals_UnboundedNotLast(t *testing.T) {
+	intervals := []PricingInterval{
+		{MinTokens: 0, MaxTokens: nil, InputPrice: testPtrFloat64(1e-6)},
+		{MinTokens: 128000, MaxTokens: testPtrInt(256000), InputPrice: testPtrFloat64(2e-6)},
+	}
+	err := ValidateIntervals(intervals)
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "unbounded")
+	require.Contains(t, err.Error(), "last")
+}
diff --git a/backend/internal/service/gateway_channel_restriction_fallback_test.go b/backend/internal/service/gateway_channel_restriction_fallback_test.go
new file mode 100644
index 00000000..d3196419
--- /dev/null
+++ b/backend/internal/service/gateway_channel_restriction_fallback_test.go
@@ -0,0 +1,130 @@
+//go:build unit
+
+package service
+
+import (
+	"context"
+	"testing"
+
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ctxkey"
+	"github.com/stretchr/testify/require"
+)
+
+func TestSelectAccountForModelWithExclusions_UsesFallbackGroupForChannelRestriction(t *testing.T) {
+	t.Parallel()
+
+	groupID := int64(10)
+	fallbackID := int64(11)
+	ch := Channel{
+		ID:             1,
+		Status:         StatusActive,
+		GroupIDs:       []int64{fallbackID},
+		RestrictModels: true,
+		ModelPricing: []ChannelModelPricing{
+			{Platform: PlatformAnthropic, Models: []string{"claude-sonnet-4-6"}},
+		},
+	}
+	channelSvc := newTestChannelService(makeStandardRepo(ch, map[int64]string{
+		fallbackID: PlatformAnthropic,
+	}))
+	accountRepo := &mockAccountRepoForPlatform{
+		accounts: []Account{
+			{ID: 1, Platform: PlatformAnthropic, Priority: 1, Status: StatusActive, Schedulable: true},
+		},
+		accountsByID: map[int64]*Account{},
+	}
+	for i := range accountRepo.accounts {
+		accountRepo.accountsByID[accountRepo.accounts[i].ID] = &accountRepo.accounts[i]
+	}
+	groupRepo := &mockGroupRepoForGateway{
+		groups: map[int64]*Group{
+			groupID: {
+				ID:              groupID,
+				Platform:        PlatformAnthropic,
+				Status:          StatusActive,
+				ClaudeCodeOnly:  true,
+				FallbackGroupID: &fallbackID,
+				Hydrated:        true,
+			},
+			fallbackID: {
+				ID:       fallbackID,
+				Platform: PlatformAnthropic,
+				Status:   StatusActive,
+				Hydrated: true,
+			},
+		},
+	}
+
+	svc := &GatewayService{
+		accountRepo:    accountRepo,
+		groupRepo:      groupRepo,
+		channelService: channelSvc,
+		cfg:            testConfig(),
+	}
+
+	ctx := context.WithValue(context.Background(), ctxkey.Group, groupRepo.groups[groupID])
+	account, err := svc.SelectAccountForModelWithExclusions(ctx, &groupID, "", "claude-sonnet-4-6", nil)
+	require.NoError(t, err)
+	require.NotNil(t, account)
+	require.Equal(t, int64(1), account.ID)
+}
+
+func TestSelectAccountWithLoadAwareness_UsesFallbackGroupForChannelRestriction(t *testing.T) {
+	t.Parallel()
+
+	groupID := int64(10)
+	fallbackID := int64(11)
+	ch := Channel{
+		ID:             1,
+		Status:         StatusActive,
+		GroupIDs:       []int64{fallbackID},
+		RestrictModels: true,
+		ModelPricing: []ChannelModelPricing{
+			{Platform: PlatformAnthropic, Models: []string{"claude-sonnet-4-6"}},
+		},
+	}
+	channelSvc := newTestChannelService(makeStandardRepo(ch, map[int64]string{
+		fallbackID: PlatformAnthropic,
+	}))
+	accountRepo := &mockAccountRepoForPlatform{
+		accounts: []Account{
+			{ID: 1, Platform: PlatformAnthropic, Priority: 1, Status: StatusActive, Schedulable: true},
+		},
+		accountsByID: map[int64]*Account{},
+	}
+	for i := range accountRepo.accounts {
+		accountRepo.accountsByID[accountRepo.accounts[i].ID] = &accountRepo.accounts[i]
+	}
+	groupRepo := &mockGroupRepoForGateway{
+		groups: map[int64]*Group{
+			groupID: {
+				ID:              groupID,
+				Platform:        PlatformAnthropic,
+				Status:          StatusActive,
+				ClaudeCodeOnly:  true,
+				FallbackGroupID: &fallbackID,
+				Hydrated:        true,
+			},
+			fallbackID: {
+				ID:       fallbackID,
+				Platform: PlatformAnthropic,
+				Status:   StatusActive,
+				Hydrated: true,
+			},
+		},
+	}
+
+	svc := &GatewayService{
+		accountRepo:    accountRepo,
+		groupRepo:      groupRepo,
+		channelService: channelSvc,
+		cfg:            testConfig(),
+	}
+
+	ctx := context.WithValue(context.Background(), ctxkey.Group, groupRepo.groups[groupID])
+	result, err := svc.SelectAccountWithLoadAwareness(ctx, &groupID, "", "claude-sonnet-4-6", nil, "", 0)
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.NotNil(t, result.Account)
+	require.Equal(t, int64(1), result.Account.ID)
+}
diff --git a/backend/internal/service/gateway_service.go b/backend/internal/service/gateway_service.go
index d272b8de..14937dd4 100644
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -1178,11 +1178,6 @@ func (s *GatewayService) SelectAccountForModel(ctx context.Context, groupID *int
 
 // SelectAccountForModelWithExclusions selects an account supporting the requested model while excluding specified accounts.
 func (s *GatewayService) SelectAccountForModelWithExclusions(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}) (*Account, error) {
-	// 渠道定价限制预检查（requested / channel_mapped 基准）
-	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
-		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
-	}
-
 	// 优先检查 context 中的强制平台（/antigravity 路由）
 	var platform string
 	forcePlatform, hasForcePlatform := ctx.Value(ctxkey.ForcePlatform).(string)
@@ -1201,6 +1196,12 @@ func (s *GatewayService) SelectAccountForModelWithExclusions(ctx context.Context
 		platform = PlatformAnthropic
 	}
 
+	// Claude Code 限制可能已将 groupID 解析为 fallback group，
+	// 渠道限制预检查必须使用解析后的分组。
+	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
+		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
+	}
+
 	// anthropic/gemini 分组支持混合调度（包含启用了 mixed_scheduling 的 antigravity 账户）
 	// 注意：强制平台模式不走混合调度
 	if (platform == PlatformAnthropic || platform == PlatformGemini) && !hasForcePlatform {
@@ -1217,11 +1218,6 @@ func (s *GatewayService) SelectAccountForModelWithExclusions(ctx context.Context
 // metadataUserID: 用于客户端亲和调度，从中提取客户端 ID
 // sub2apiUserID: 系统用户 ID，用于二维亲和调度
 func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}, metadataUserID string, sub2apiUserID int64) (*AccountSelectionResult, error) {
-	// 渠道定价限制预检查（requested / channel_mapped 基准）
-	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
-		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
-	}
-
 	// 调试日志：记录调度入口参数
 	excludedIDsList := make([]int64, 0, len(excludedIDs))
 	for id := range excludedIDs {
@@ -1242,6 +1238,12 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 	}
 	ctx = s.withGroupContext(ctx, group)
 
+	// Claude Code 限制可能已将 groupID 解析为 fallback group，
+	// 渠道限制预检查必须使用解析后的分组。
+	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
+		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
+	}
+
 	var stickyAccountID int64
 	if prefetch := prefetchedStickyAccountIDFromContext(ctx, groupID); prefetch > 0 {
 		stickyAccountID = prefetch
diff --git a/backend/internal/service/openai_channel_restriction_test.go b/backend/internal/service/openai_channel_restriction_test.go
new file mode 100644
index 00000000..c9dbceab
--- /dev/null
+++ b/backend/internal/service/openai_channel_restriction_test.go
@@ -0,0 +1,140 @@
+//go:build unit
+
+package service
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestOpenAISelectAccountForModelWithExclusions_ChannelMappedRestrictionRejectsEarly(t *testing.T) {
+	t.Parallel()
+
+	channelSvc := newTestChannelService(makeStandardRepo(Channel{
+		ID:                 1,
+		Status:             StatusActive,
+		GroupIDs:           []int64{10},
+		RestrictModels:     true,
+		BillingModelSource: BillingModelSourceChannelMapped,
+		ModelPricing: []ChannelModelPricing{
+			{Platform: PlatformOpenAI, Models: []string{"gpt-4o"}},
+		},
+		ModelMapping: map[string]map[string]string{
+			PlatformOpenAI: {"gpt-4.1": "o3-mini"},
+		},
+	}, map[int64]string{10: PlatformOpenAI}))
+
+	svc := &OpenAIGatewayService{
+		accountRepo: stubOpenAIAccountRepo{accounts: []Account{
+			{ID: 1, Platform: PlatformOpenAI, Status: StatusActive, Schedulable: true},
+		}},
+		channelService: channelSvc,
+	}
+
+	groupID := int64(10)
+	_, err := svc.SelectAccountForModelWithExclusions(context.Background(), &groupID, "", "gpt-4.1", nil)
+	require.ErrorIs(t, err, ErrNoAvailableAccounts)
+	require.Contains(t, err.Error(), "channel pricing restriction")
+}
+
+func TestOpenAISelectAccountForModelWithExclusions_UpstreamRestrictionSkipsDisallowedAccount(t *testing.T) {
+	t.Parallel()
+
+	channelSvc := newTestChannelService(makeStandardRepo(Channel{
+		ID:                 1,
+		Status:             StatusActive,
+		GroupIDs:           []int64{10},
+		RestrictModels:     true,
+		BillingModelSource: BillingModelSourceUpstream,
+		ModelPricing: []ChannelModelPricing{
+			{Platform: PlatformOpenAI, Models: []string{"o3-mini"}},
+		},
+	}, map[int64]string{10: PlatformOpenAI}))
+
+	svc := &OpenAIGatewayService{
+		accountRepo: stubOpenAIAccountRepo{accounts: []Account{
+			{
+				ID:          1,
+				Platform:    PlatformOpenAI,
+				Status:      StatusActive,
+				Schedulable: true,
+				Priority:    10,
+				Credentials: map[string]any{
+					"model_mapping": map[string]any{"gpt-4.1": "gpt-4o"},
+				},
+			},
+			{
+				ID:          2,
+				Platform:    PlatformOpenAI,
+				Status:      StatusActive,
+				Schedulable: true,
+				Priority:    20,
+				Credentials: map[string]any{
+					"model_mapping": map[string]any{"gpt-4.1": "o3-mini"},
+				},
+			},
+		}},
+		channelService: channelSvc,
+	}
+
+	groupID := int64(10)
+	account, err := svc.SelectAccountForModelWithExclusions(context.Background(), &groupID, "", "gpt-4.1", nil)
+	require.NoError(t, err)
+	require.NotNil(t, account)
+	require.Equal(t, int64(2), account.ID)
+}
+
+func TestOpenAISelectAccountForModelWithExclusions_StickyRestrictedUpstreamFallsBack(t *testing.T) {
+	t.Parallel()
+
+	channelSvc := newTestChannelService(makeStandardRepo(Channel{
+		ID:                 1,
+		Status:             StatusActive,
+		GroupIDs:           []int64{10},
+		RestrictModels:     true,
+		BillingModelSource: BillingModelSourceUpstream,
+		ModelPricing: []ChannelModelPricing{
+			{Platform: PlatformOpenAI, Models: []string{"o3-mini"}},
+		},
+	}, map[int64]string{10: PlatformOpenAI}))
+
+	cache := &stubGatewayCache{
+		sessionBindings: map[string]int64{"openai:sticky-session": 1},
+	}
+	svc := &OpenAIGatewayService{
+		accountRepo: stubOpenAIAccountRepo{accounts: []Account{
+			{
+				ID:          1,
+				Platform:    PlatformOpenAI,
+				Status:      StatusActive,
+				Schedulable: true,
+				Priority:    10,
+				Credentials: map[string]any{
+					"model_mapping": map[string]any{"gpt-4.1": "gpt-4o"},
+				},
+			},
+			{
+				ID:          2,
+				Platform:    PlatformOpenAI,
+				Status:      StatusActive,
+				Schedulable: true,
+				Priority:    20,
+				Credentials: map[string]any{
+					"model_mapping": map[string]any{"gpt-4.1": "o3-mini"},
+				},
+			},
+		}},
+		channelService: channelSvc,
+		cache:          cache,
+	}
+
+	groupID := int64(10)
+	account, err := svc.SelectAccountForModelWithExclusions(context.Background(), &groupID, "sticky-session", "gpt-4.1", nil)
+	require.NoError(t, err)
+	require.NotNil(t, account)
+	require.Equal(t, int64(2), account.ID)
+	require.Equal(t, 1, cache.deletedSessions["openai:sticky-session"])
+	require.Equal(t, int64(2), cache.sessionBindings["openai:sticky-session"])
+}
diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go
index ef0aaa5b..d1c52ea5 100644
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -10,6 +10,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"log/slog"
 	"math/rand"
 	"net/http"
 	"sort"
@@ -423,6 +424,44 @@ func (s *OpenAIGatewayService) ResolveChannelMappingAndRestrict(ctx context.Cont
 	return s.channelService.ResolveChannelMappingAndRestrict(ctx, groupID, model)
 }
 
+func (s *OpenAIGatewayService) checkChannelPricingRestriction(ctx context.Context, groupID *int64, requestedModel string) bool {
+	if groupID == nil || s.channelService == nil || requestedModel == "" {
+		return false
+	}
+	mapping := s.channelService.ResolveChannelMapping(ctx, *groupID, requestedModel)
+	billingModel := billingModelForRestriction(mapping.BillingModelSource, requestedModel, mapping.MappedModel)
+	if billingModel == "" {
+		return false
+	}
+	return s.channelService.IsModelRestricted(ctx, *groupID, billingModel)
+}
+
+func (s *OpenAIGatewayService) isUpstreamModelRestrictedByChannel(ctx context.Context, groupID int64, account *Account, requestedModel string) bool {
+	if s.channelService == nil {
+		return false
+	}
+	upstreamModel := resolveOpenAIForwardModel(account, requestedModel, "")
+	if upstreamModel == "" {
+		return false
+	}
+	return s.channelService.IsModelRestricted(ctx, groupID, upstreamModel)
+}
+
+func (s *OpenAIGatewayService) needsUpstreamChannelRestrictionCheck(ctx context.Context, groupID *int64) bool {
+	if groupID == nil || s.channelService == nil {
+		return false
+	}
+	ch, err := s.channelService.GetChannelForGroup(ctx, *groupID)
+	if err != nil {
+		slog.Warn("failed to check openai channel upstream restriction", "group_id", *groupID, "error", err)
+		return false
+	}
+	if ch == nil || !ch.RestrictModels {
+		return false
+	}
+	return ch.BillingModelSource == BillingModelSourceUpstream
+}
+
 // ReplaceModelInBody 替换请求体中的 JSON model 字段（通用 gjson/sjson 实现）。
 func (s *OpenAIGatewayService) ReplaceModelInBody(body []byte, newModel string) []byte {
 	return ReplaceModelInBody(body, newModel)
@@ -1162,6 +1201,10 @@ func (s *OpenAIGatewayService) SelectAccountForModelWithExclusions(ctx context.C
 }
 
 func (s *OpenAIGatewayService) selectAccountForModelWithExclusions(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}, stickyAccountID int64) (*Account, error) {
+	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
+		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
+	}
+
 	// 1. 尝试粘性会话命中
 	// Try sticky session hit
 	if account := s.tryStickySessionHit(ctx, groupID, sessionHash, requestedModel, excludedIDs, stickyAccountID); account != nil {
@@ -1177,7 +1220,7 @@ func (s *OpenAIGatewayService) selectAccountForModelWithExclusions(ctx context.C
 
 	// 3. 按优先级 + LRU 选择最佳账号
 	// Select by priority + LRU
-	selected := s.selectBestAccount(ctx, accounts, requestedModel, excludedIDs)
+	selected := s.selectBestAccount(ctx, groupID, accounts, requestedModel, excludedIDs)
 
 	if selected == nil {
 		if requestedModel != "" {
@@ -1243,6 +1286,11 @@ func (s *OpenAIGatewayService) tryStickySessionHit(ctx context.Context, groupID
 		_ = s.deleteStickySessionAccountID(ctx, groupID, sessionHash)
 		return nil
 	}
+	if groupID != nil && s.needsUpstreamChannelRestrictionCheck(ctx, groupID) &&
+		s.isUpstreamModelRestrictedByChannel(ctx, *groupID, account, requestedModel) {
+		_ = s.deleteStickySessionAccountID(ctx, groupID, sessionHash)
+		return nil
+	}
 
 	// 刷新会话 TTL 并返回账号
 	// Refresh session TTL and return account
@@ -1255,8 +1303,9 @@ func (s *OpenAIGatewayService) tryStickySessionHit(ctx context.Context, groupID
 //
 // selectBestAccount selects the best account from candidates (priority + LRU).
 // Returns nil if no available account.
-func (s *OpenAIGatewayService) selectBestAccount(ctx context.Context, accounts []Account, requestedModel string, excludedIDs map[int64]struct{}) *Account {
+func (s *OpenAIGatewayService) selectBestAccount(ctx context.Context, groupID *int64, accounts []Account, requestedModel string, excludedIDs map[int64]struct{}) *Account {
 	var selected *Account
+	needsUpstreamCheck := s.needsUpstreamChannelRestrictionCheck(ctx, groupID)
 
 	for i := range accounts {
 		acc := &accounts[i]
@@ -1275,6 +1324,9 @@ func (s *OpenAIGatewayService) selectBestAccount(ctx context.Context, accounts [
 		if fresh == nil {
 			continue
 		}
+		if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, fresh, requestedModel) {
+			continue
+		}
 
 		// 选择优先级最高且最久未使用的账号
 		// Select highest priority and least recently used
@@ -1326,7 +1378,12 @@ func (s *OpenAIGatewayService) isBetterAccount(candidate, current *Account) bool
 
 // SelectAccountWithLoadAwareness selects an account with load-awareness and wait plan.
 func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}) (*AccountSelectionResult, error) {
+	if s.checkChannelPricingRestriction(ctx, groupID, requestedModel) {
+		return nil, fmt.Errorf("%w supporting model: %s (channel pricing restriction)", ErrNoAvailableAccounts, requestedModel)
+	}
+
 	cfg := s.schedulingConfig()
+	needsUpstreamCheck := s.needsUpstreamChannelRestrictionCheck(ctx, groupID)
 	var stickyAccountID int64
 	if sessionHash != "" && s.cache != nil {
 		if accountID, err := s.getStickySessionAccountID(ctx, groupID, sessionHash); err == nil {
@@ -1402,6 +1459,8 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 					account = s.recheckSelectedOpenAIAccountFromDB(ctx, account, requestedModel)
 					if account == nil {
 						_ = s.deleteStickySessionAccountID(ctx, groupID, sessionHash)
+					} else if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, account, requestedModel) {
+						_ = s.deleteStickySessionAccountID(ctx, groupID, sessionHash)
 					} else {
 						result, err := s.tryAcquireAccountSlot(ctx, accountID, account.Concurrency)
 						if err == nil && result.Acquired {
@@ -1447,6 +1506,9 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 		if requestedModel != "" && !acc.IsModelSupported(requestedModel) {
 			continue
 		}
+		if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, acc, requestedModel) {
+			continue
+		}
 		candidates = append(candidates, acc)
 	}
 
@@ -1471,6 +1533,9 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 			if fresh == nil {
 				continue
 			}
+			if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, fresh, requestedModel) {
+				continue
+			}
 			result, err := s.tryAcquireAccountSlot(ctx, fresh.ID, fresh.Concurrency)
 			if err == nil && result.Acquired {
 				if sessionHash != "" {
@@ -1525,6 +1590,9 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 				if fresh == nil {
 					continue
 				}
+				if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, fresh, requestedModel) {
+					continue
+				}
 				result, err := s.tryAcquireAccountSlot(ctx, fresh.ID, fresh.Concurrency)
 				if err == nil && result.Acquired {
 					if sessionHash != "" {
@@ -1547,6 +1615,9 @@ func (s *OpenAIGatewayService) SelectAccountWithLoadAwareness(ctx context.Contex
 		if fresh == nil {
 			continue
 		}
+		if needsUpstreamCheck && s.isUpstreamModelRestrictedByChannel(ctx, *groupID, fresh, requestedModel) {
+			continue
+		}
 		return &AccountSelectionResult{
 			Account: fresh,
 			WaitPlan: &AccountWaitPlan{
diff --git a/frontend/src/components/admin/channel/types.ts b/frontend/src/components/admin/channel/types.ts
index 431e3eeb..8d998911 100644
--- a/frontend/src/components/admin/channel/types.ts
+++ b/frontend/src/components/admin/channel/types.ts
@@ -113,6 +113,70 @@ export function findModelConflict(models: string[]): [string, string] | null {
   return null
 }
 
+// ── 区间校验 ──────────────────────────────────────────────
+
+/** 校验区间列表的合法性，返回错误消息；通过则返回 null */
+export function validateIntervals(intervals: IntervalFormEntry[]): string | null {
+  if (!intervals || intervals.length === 0) return null
+
+  // 按 min_tokens 排序（不修改原数组）
+  const sorted = [...intervals].sort((a, b) => a.min_tokens - b.min_tokens)
+
+  for (let i = 0; i < sorted.length; i++) {
+    const err = validateSingleInterval(sorted[i], i)
+    if (err) return err
+  }
+  return checkIntervalOverlap(sorted)
+}
+
+function validateSingleInterval(iv: IntervalFormEntry, idx: number): string | null {
+  if (iv.min_tokens < 0) {
+    return `区间 #${idx + 1}: 最小 token 数 (${iv.min_tokens}) 不能为负数`
+  }
+  if (iv.max_tokens != null) {
+    if (iv.max_tokens <= 0) {
+      return `区间 #${idx + 1}: 最大 token 数 (${iv.max_tokens}) 必须大于 0`
+    }
+    if (iv.max_tokens <= iv.min_tokens) {
+      return `区间 #${idx + 1}: 最大 token 数 (${iv.max_tokens}) 必须大于最小 token 数 (${iv.min_tokens})`
+    }
+  }
+  return validateIntervalPrices(iv, idx)
+}
+
+function validateIntervalPrices(iv: IntervalFormEntry, idx: number): string | null {
+  const prices: [string, number | string | null][] = [
+    ['输入价格', iv.input_price],
+    ['输出价格', iv.output_price],
+    ['缓存写入价格', iv.cache_write_price],
+    ['缓存读取价格', iv.cache_read_price],
+    ['单次价格', iv.per_request_price],
+  ]
+  for (const [name, val] of prices) {
+    if (val != null && val !== '' && Number(val) < 0) {
+      return `区间 #${idx + 1}: ${name}不能为负数`
+    }
+  }
+  return null
+}
+
+function checkIntervalOverlap(sorted: IntervalFormEntry[]): string | null {
+  for (let i = 0; i < sorted.length; i++) {
+    // 无上限区间必须是最后一个
+    if (sorted[i].max_tokens == null && i < sorted.length - 1) {
+      return `区间 #${i + 1}: 无上限区间（最大 token 数为空）只能是最后一个`
+    }
+    if (i === 0) continue
+    const prev = sorted[i - 1]
+    // (min, max] 语义：前一个区间上界 > 当前区间下界则重叠
+    if (prev.max_tokens == null || prev.max_tokens > sorted[i].min_tokens) {
+      const prevMax = prev.max_tokens == null ? '∞' : String(prev.max_tokens)
+      return `区间 #${i} 和 #${i + 1} 重叠：前一个区间上界 (${prevMax}) 大于当前区间下界 (${sorted[i].min_tokens})`
+    }
+  }
+  return null
+}
+
 /** 平台对应的模型 tag 样式（背景+文字） */
 export function getPlatformTagClass(platform: string): string {
   switch (platform) {
diff --git a/frontend/src/views/admin/ChannelsView.vue b/frontend/src/views/admin/ChannelsView.vue
index 4d4150fb..b651be7d 100644
--- a/frontend/src/views/admin/ChannelsView.vue
+++ b/frontend/src/views/admin/ChannelsView.vue
@@ -418,7 +418,7 @@ import { useAppStore } from '@/stores/app'
 import { adminAPI } from '@/api/admin'
 import type { Channel, ChannelModelPricing, CreateChannelRequest, UpdateChannelRequest } from '@/api/admin/channels'
 import type { PricingFormEntry } from '@/components/admin/channel/types'
-import { mTokToPerToken, perTokenToMTok, apiIntervalsToForm, formIntervalsToAPI, findModelConflict } from '@/components/admin/channel/types'
+import { mTokToPerToken, perTokenToMTok, apiIntervalsToForm, formIntervalsToAPI, findModelConflict, validateIntervals } from '@/components/admin/channel/types'
 import type { AdminGroup, GroupPlatform } from '@/types'
 import type { Column } from '@/components/common/types'
 import AppLayout from '@/components/layout/AppLayout.vue'
@@ -922,6 +922,21 @@ async function handleSubmit() {
     }
   }
 
+  // 校验区间合法性（范围、重叠等）
+  for (const section of form.platforms.filter(s => s.enabled)) {
+    for (const entry of section.model_pricing) {
+      if (!entry.intervals || entry.intervals.length === 0) continue
+      const intervalErr = validateIntervals(entry.intervals)
+      if (intervalErr) {
+        const platformLabel = t('admin.groups.platforms.' + section.platform, section.platform)
+        const modelLabel = entry.models.join(', ') || '未命名'
+        appStore.showError(`${platformLabel} - ${modelLabel}: ${intervalErr}`)
+        activeTab.value = section.platform
+        return
+      }
+    }
+  }
+
   const { group_ids, model_pricing, model_mapping } = formToAPI()
 
   submitting.value = true