diff --git a/backend/cmd/server/wire_gen.go b/backend/cmd/server/wire_gen.go
index b89353fc9e..1d39fa1e31 100644
--- a/backend/cmd/server/wire_gen.go
+++ b/backend/cmd/server/wire_gen.go
@@ -143,7 +143,7 @@ func initializeApplication(buildInfo handler.BuildInfo) (*Application, error) {
 	schedulerSnapshotService := service.ProvideSchedulerSnapshotService(schedulerCache, schedulerOutboxRepository, accountRepository, groupRepository, configConfig)
 	antigravityTokenProvider := service.ProvideAntigravityTokenProvider(accountRepository, geminiTokenCache, antigravityOAuthService, oAuthRefreshAPI, tempUnschedCache)
 	internal500CounterCache := repository.NewInternal500CounterCache(redisClient)
-	antigravityGatewayService := service.NewAntigravityGatewayService(accountRepository, gatewayCache, schedulerSnapshotService, antigravityTokenProvider, rateLimitService, httpUpstream, settingService, internal500CounterCache, accountUsageService)
+	antigravityGatewayService := service.NewAntigravityGatewayService(accountRepository, gatewayCache, schedulerSnapshotService, antigravityTokenProvider, rateLimitService, httpUpstream, settingService, internal500CounterCache)
 	accountTestService := service.NewAccountTestService(accountRepository, geminiTokenProvider, antigravityGatewayService, httpUpstream, configConfig, tlsFingerprintProfileService)
 	crsSyncService := service.NewCRSSyncService(accountRepository, proxyRepository, oAuthService, openAIOAuthService, geminiOAuthService, configConfig)
 	accountHandler := admin.NewAccountHandler(adminService, oAuthService, openAIOAuthService, geminiOAuthService, antigravityOAuthService, rateLimitService, accountUsageService, accountTestService, concurrencyService, crsSyncService, sessionLimitCache, rpmCache, compositeTokenCacheInvalidator)
diff --git a/backend/internal/domain/constants.go b/backend/internal/domain/constants.go
index 8168f7ec0e..429486c3bf 100644
--- a/backend/internal/domain/constants.go
+++ b/backend/internal/domain/constants.go
@@ -75,11 +75,11 @@ var DefaultAntigravityModelMapping = map[string]string{
 	"claude-opus-4-6":            "claude-opus-4-6-thinking", // 简称映射
 	"claude-opus-4-5-thinking":   "claude-opus-4-6-thinking", // 迁移旧模型
 	"claude-sonnet-4-6":          "claude-sonnet-4-6",
-	"claude-sonnet-4-5":          "claude-sonnet-4-6", // Sonnet 4.5 → 4.6
-	"claude-sonnet-4-5-thinking": "claude-sonnet-4-6", // Sonnet 4.5 → 4.6
+	"claude-sonnet-4-5":          "claude-sonnet-4-5",
+	"claude-sonnet-4-5-thinking": "claude-sonnet-4-5-thinking",
 	// Claude 详细版本 ID 映射
 	"claude-opus-4-5-20251101":   "claude-opus-4-6-thinking", // 迁移旧模型
-	"claude-sonnet-4-5-20250929": "claude-sonnet-4-6",        // Sonnet 4.5 → 4.6
+	"claude-sonnet-4-5-20250929": "claude-sonnet-4-5",
 	// Claude Haiku → Sonnet（无 Haiku 支持）
 	"claude-haiku-4-5":          "claude-sonnet-4-6",
 	"claude-haiku-4-5-20251001": "claude-sonnet-4-6",
diff --git a/backend/internal/service/account_usage_service.go b/backend/internal/service/account_usage_service.go
index 50deba8bd0..0e5741d818 100644
--- a/backend/internal/service/account_usage_service.go
+++ b/backend/internal/service/account_usage_service.go
@@ -846,22 +846,6 @@ func (s *AccountUsageService) getAntigravityUsage(ctx context.Context, account *
 	return usage, nil
 }
 
-// GetAntigravityCredits 返回账号的 AI Credits 信息，复用 getAntigravityUsage 的缓存。
-// 如果缓存存在且 TTL 充足则直接返回；TTL 不足时自动刷新。
-func (s *AccountUsageService) GetAntigravityCredits(ctx context.Context, account *Account) (*UsageInfo, error) {
-	if account == nil || account.Platform != PlatformAntigravity {
-		return nil, nil
-	}
-	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 e1b787415d..ec3650859c 100644
--- a/backend/internal/service/antigravity_credits_overages.go
+++ b/backend/internal/service/antigravity_credits_overages.go
@@ -4,7 +4,6 @@ import (
 	"context"
 	"encoding/json"
 	"io"
-	"log/slog"
 	"net/http"
 	"strings"
 	"time"
@@ -18,140 +17,8 @@ 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。
-// 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 {
-		slog.Error("check_credits: get_credits_failed",
-			"account_id", account.ID, "error", err)
-		return true // 出错时 fail-open
-	}
-
-	// 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 {
-		slog.Warn("check_credits: insufficient credits",
-			"account_id", account.ID)
-	}
-	return hasCredits
-}
-
-// hasEnoughCredits 检查 UsageInfo 中是否有足够的 GOOGLE_ONE_AI 积分。
-// 返回 true 表示积分可用，false 表示积分不足或无积分信息。
-func hasEnoughCredits(info *UsageInfo) bool {
-	if info == nil || len(info.AICredits) == 0 {
-		return false
-	}
-
-	for _, credit := range info.AICredits {
-		if credit.CreditType == "GOOGLE_ONE_AI" {
-			minimum := credit.MinimumBalance
-			if minimum <= 0 {
-				minimum = 5
-			}
-			return credit.Amount >= minimum
-		}
-	}
-
-	return false
-}
-
 type antigravity429Category string
 
 const (
@@ -224,10 +91,6 @@ func (s *AntigravityGatewayService) clearCreditsExhausted(ctx context.Context, a
 	}); err != nil {
 		logger.LegacyPrintf("service.antigravity_gateway", "clear credits exhausted failed: account=%d err=%v", account.ID, err)
 	}
-	// 同步更新 Redis 调度快照，避免其他节点/请求延迟感知
-	if s.schedulerSnapshot != nil {
-		_ = s.schedulerSnapshot.UpdateAccountInCache(ctx, account)
-	}
 }
 
 // classifyAntigravity429 将 Antigravity 的 429 响应归类为配额耗尽、限流或未知。
@@ -278,14 +141,6 @@ func resolveCreditsOveragesModelKey(ctx context.Context, account *Account, upstr
 }
 
 // shouldMarkCreditsExhausted 判断一次 credits 请求失败是否应标记为 credits 耗尽。
-// 此函数在积分注入后失败时调用（预检查注入 + attemptCreditsOveragesRetry 两条路径）。
-//   - 429 + 单模型限流（"exhausted your capacity on this model"）：该模型免费配额用完，
-//     积分注入对此无效，但账号积分对其他模型可能仍可用 → 不标记积分耗尽。
-//   - 429 + 积分关键词（如 "insufficient credits"）：积分确实不足 → 标记耗尽。
-//   - 429 + 无积分关键词（节点限流、瞬时 rate limit 等）：与积分无关 → 不标记。
-//   - 403 等其他 4xx：检查 body 是否包含积分不足的关键词。
-//
-// clearCreditsExhausted 会在后续成功时自动清除。
 func shouldMarkCreditsExhausted(resp *http.Response, respBody []byte, reqErr error) bool {
 	if reqErr != nil || resp == nil {
 		return false
@@ -293,23 +148,13 @@ func shouldMarkCreditsExhausted(resp *http.Response, respBody []byte, reqErr err
 	if resp.StatusCode >= 500 || resp.StatusCode == http.StatusRequestTimeout {
 		return false
 	}
-	bodyLower := strings.ToLower(string(respBody))
-	// 积分注入后仍 429
-	if resp.StatusCode == http.StatusTooManyRequests {
-		// 单模型配额耗尽：积分注入对此无效，不标记整个账号积分耗尽
-		if strings.Contains(bodyLower, "exhausted your capacity on this model") {
-			return false
-		}
-		// 仅当 body 包含积分相关关键词时才标记耗尽；
-		// 节点限流、瞬时 rate limit 等非积分 429 不应触发 5 小时熔断。
-		return containsCreditsExhaustedKeyword(bodyLower)
+	// 注意：不再检查 isURLLevelRateLimit。此函数仅在积分重试失败后调用，
+	// 如果注入 enabledCreditTypes 后仍返回 "Resource has been exhausted"，
+	// 说明积分也已耗尽，应该标记。clearCreditsExhausted 会在后续成功时自动清除。
+	if info := parseAntigravitySmartRetryInfo(respBody); info != nil {
+		return false
 	}
-	// 其他 4xx：关键词匹配（如 403 + "Insufficient credits"）
-	return containsCreditsExhaustedKeyword(bodyLower)
-}
-
-// containsCreditsExhaustedKeyword 检查响应体是否包含积分耗尽的关键词。
-func containsCreditsExhaustedKeyword(bodyLower string) bool {
+	bodyLower := strings.ToLower(string(respBody))
 	for _, keyword := range creditsExhaustedKeywords {
 		if strings.Contains(bodyLower, keyword) {
 			return true
@@ -336,16 +181,6 @@ func (s *AntigravityGatewayService) attemptCreditsOveragesRetry(
 	if creditsBody == nil {
 		return &creditsOveragesRetryResult{handled: false}
 	}
-
-	// Check actual credits balance before attempting retry
-	if !s.checkAccountCredits(p.ctx, p.account) {
-		s.setCreditsExhausted(p.ctx, p.account)
-		modelKey := resolveCreditsOveragesModelKey(p.ctx, p.account, modelName, p.requestedModel)
-		logger.LegacyPrintf("service.antigravity_gateway", "%s credit_overages_no_credits model=%s account=%d (skipping credits retry)",
-			p.prefix, modelKey, p.account.ID)
-		return &creditsOveragesRetryResult{handled: true}
-	}
-
 	modelKey := resolveCreditsOveragesModelKey(p.ctx, p.account, modelName, p.requestedModel)
 	logger.LegacyPrintf("service.antigravity_gateway", "%s status=429 credit_overages_retry model=%s account=%d (injecting enabledCreditTypes)",
 		p.prefix, modelKey, p.account.ID)
diff --git a/backend/internal/service/antigravity_credits_overages_test.go b/backend/internal/service/antigravity_credits_overages_test.go
index 1e9106b41d..7a5224daf3 100644
--- a/backend/internal/service/antigravity_credits_overages_test.go
+++ b/backend/internal/service/antigravity_credits_overages_test.go
@@ -239,7 +239,7 @@ func TestAntigravityRetryLoop_ModelRateLimited_InjectsCredits(t *testing.T) {
 		Extra: map[string]any{
 			"allow_overages": true,
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limited_at":     time.Now().UTC().Format(time.RFC3339),
 					"rate_limit_reset_at": time.Now().Add(30 * time.Minute).UTC().Format(time.RFC3339),
 				},
@@ -290,7 +290,7 @@ func TestAntigravityRetryLoop_CreditsExhausted_DoesNotInject(t *testing.T) {
 		Extra: map[string]any{
 			"allow_overages": true,
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limited_at":     time.Now().UTC().Format(time.RFC3339),
 					"rate_limit_reset_at": time.Now().Add(30 * time.Minute).UTC().Format(time.RFC3339),
 				},
@@ -355,7 +355,7 @@ func TestAntigravityRetryLoop_CreditErrorMarksExhausted(t *testing.T) {
 		Extra: map[string]any{
 			"allow_overages": true,
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limited_at":     time.Now().UTC().Format(time.RFC3339),
 					"rate_limit_reset_at": time.Now().Add(30 * time.Minute).UTC().Format(time.RFC3339),
 				},
@@ -418,25 +418,13 @@ func TestShouldMarkCreditsExhausted(t *testing.T) {
 		require.True(t, shouldMarkCreditsExhausted(resp, body, nil))
 	})
 
-	t.Run("单模型配额耗尽不标记（积分对此无效）", func(t *testing.T) {
-		resp := &http.Response{StatusCode: http.StatusTooManyRequests}
-		body := []byte(`{"error":{"code":429,"message":"You have exhausted your capacity on this model. Your quota will reset after 146h11m17s.","status":"RESOURCE_EXHAUSTED"}}`)
-		require.False(t, shouldMarkCreditsExhausted(resp, body, nil))
-	})
-
-	t.Run("429 结构化限流不标记（节点限流/瞬时 rate limit 与积分无关）", func(t *testing.T) {
+	t.Run("结构化限流不标记", func(t *testing.T) {
 		resp := &http.Response{StatusCode: http.StatusTooManyRequests}
 		body := []byte(`{"error":{"status":"RESOURCE_EXHAUSTED","details":[{"@type":"type.googleapis.com/google.rpc.ErrorInfo","reason":"RATE_LIMIT_EXCEEDED"},{"@type":"type.googleapis.com/google.rpc.RetryInfo","retryDelay":"0.5s"}]}}`)
 		require.False(t, shouldMarkCreditsExhausted(resp, body, nil))
 	})
 
-	t.Run("429 含积分关键词时标记", func(t *testing.T) {
-		resp := &http.Response{StatusCode: http.StatusTooManyRequests}
-		body := []byte(`{"error":{"message":"Insufficient credits to complete this request"}}`)
-		require.True(t, shouldMarkCreditsExhausted(resp, body, nil))
-	})
-
-	t.Run("含 credits 关键词时标记（403 场景）", func(t *testing.T) {
+	t.Run("含 credits 关键词时标记", func(t *testing.T) {
 		resp := &http.Response{StatusCode: http.StatusForbidden}
 		for _, keyword := range []string{
 			"Insufficient GOOGLE_ONE_AI credits",
@@ -515,7 +503,7 @@ func TestClearCreditsExhausted(t *testing.T) {
 			ID: 1,
 			Extra: map[string]any{
 				modelRateLimitsKey: map[string]any{
-					"claude-sonnet-4-6": map[string]any{
+					"claude-sonnet-4-5": map[string]any{
 						"rate_limited_at":     "2026-03-15T00:00:00Z",
 						"rate_limit_reset_at": "2099-03-15T00:00:00Z",
 					},
@@ -532,7 +520,7 @@ func TestClearCreditsExhausted(t *testing.T) {
 			ID: 1,
 			Extra: map[string]any{
 				modelRateLimitsKey: map[string]any{
-					"claude-sonnet-4-6": map[string]any{
+					"claude-sonnet-4-5": map[string]any{
 						"rate_limited_at":     "2026-03-15T00:00:00Z",
 						"rate_limit_reset_at": "2099-03-15T00:00:00Z",
 					},
@@ -550,109 +538,7 @@ func TestClearCreditsExhausted(t *testing.T) {
 		_, exists := rawLimits[creditsExhaustedKey]
 		require.False(t, exists, "AICredits key 应被删除")
 		// 普通模型限流应保留
-		_, exists = rawLimits["claude-sonnet-4-6"]
+		_, exists = rawLimits["claude-sonnet-4-5"]
 		require.True(t, exists, "普通模型限流应保留")
 	})
 }
-
-// ===========================================================================
-// hasEnoughCredits — standalone credits balance check
-// ===========================================================================
-
-func TestHasEnoughCredits(t *testing.T) {
-	tests := []struct {
-		name string
-		info *UsageInfo
-		want bool
-	}{
-		{
-			name: "nil UsageInfo",
-			info: nil,
-			want: false,
-		},
-		{
-			name: "empty AICredits list",
-			info: &UsageInfo{AICredits: []AICredit{}},
-			want: false,
-		},
-		{
-			name: "GOOGLE_ONE_AI with enough credits (amount=18778, minimum=50)",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 18778, MinimumBalance: 50},
-				},
-			},
-			want: true,
-		},
-		{
-			name: "GOOGLE_ONE_AI below minimum (amount=3, minimum=5)",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 3, MinimumBalance: 5},
-				},
-			},
-			want: false,
-		},
-		{
-			name: "GOOGLE_ONE_AI with zero MinimumBalance defaults to 5, amount=6",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 6, MinimumBalance: 0},
-				},
-			},
-			want: true,
-		},
-		{
-			name: "GOOGLE_ONE_AI with zero MinimumBalance defaults to 5, amount=4",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 4, MinimumBalance: 0},
-				},
-			},
-			want: false,
-		},
-		{
-			name: "GOOGLE_ONE_AI exactly at minimum (amount=5, minimum=5)",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 5, MinimumBalance: 5},
-				},
-			},
-			want: true,
-		},
-		{
-			name: "no GOOGLE_ONE_AI credit type",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "OTHER_CREDIT", Amount: 10000, MinimumBalance: 5},
-				},
-			},
-			want: false,
-		},
-		{
-			name: "multiple credits, GOOGLE_ONE_AI present with enough",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "OTHER_CREDIT", Amount: 0, MinimumBalance: 5},
-					{CreditType: "GOOGLE_ONE_AI", Amount: 100, MinimumBalance: 10},
-				},
-			},
-			want: true,
-		},
-		{
-			name: "negative MinimumBalance defaults to 5",
-			info: &UsageInfo{
-				AICredits: []AICredit{
-					{CreditType: "GOOGLE_ONE_AI", Amount: 6, MinimumBalance: -1},
-				},
-			},
-			want: true,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			require.Equal(t, tt.want, hasEnoughCredits(tt.info))
-		})
-	}
-}
diff --git a/backend/internal/service/antigravity_gateway_service.go b/backend/internal/service/antigravity_gateway_service.go
index 6dba547cfc..a76e59fbea 100644
--- a/backend/internal/service/antigravity_gateway_service.go
+++ b/backend/internal/service/antigravity_gateway_service.go
@@ -557,13 +557,7 @@ func (s *AntigravityGatewayService) antigravityRetryLoop(p antigravityRetryLoopP
 	if p.requestedModel != "" && p.account.Platform == PlatformAntigravity &&
 		p.account.IsOveragesEnabled() && !p.account.isCreditsExhausted() &&
 		p.account.isModelRateLimitedWithContext(p.ctx, p.requestedModel) {
-		// Check actual credits balance before injection
-		if !s.checkAccountCredits(p.ctx, p.account) {
-			// No credits available - mark as exhausted and skip injection
-			s.setCreditsExhausted(p.ctx, p.account)
-			logger.LegacyPrintf("service.antigravity_gateway", "%s pre_check: no_credits_available account=%d (skipping credits injection)",
-				p.prefix, p.account.ID)
-		} else if creditsBody := injectEnabledCreditTypes(p.body); creditsBody != nil {
+		if creditsBody := injectEnabledCreditTypes(p.body); creditsBody != nil {
 			p.body = creditsBody
 			overagesInjected = true
 			logger.LegacyPrintf("service.antigravity_gateway", "%s pre_check: model_rate_limited_credits_inject model=%s account=%d (injecting enabledCreditTypes)",
@@ -876,15 +870,14 @@ func logPrefix(sessionID, accountName string) string {
 
 // AntigravityGatewayService 处理 Antigravity 平台的 API 转发
 type AntigravityGatewayService struct {
-	accountRepo         AccountRepository
-	tokenProvider       *AntigravityTokenProvider
-	rateLimitService    *RateLimitService
-	httpUpstream        HTTPUpstream
-	settingService      *SettingService
-	cache               GatewayCache // 用于模型级限流时清除粘性会话绑定
-	schedulerSnapshot   *SchedulerSnapshotService
-	internal500Cache    Internal500CounterCache // INTERNAL 500 渐进惩罚计数器
-	accountUsageService *AccountUsageService    // 共享 usage 缓存，用于积分余额检查
+	accountRepo       AccountRepository
+	tokenProvider     *AntigravityTokenProvider
+	rateLimitService  *RateLimitService
+	httpUpstream      HTTPUpstream
+	settingService    *SettingService
+	cache             GatewayCache // 用于模型级限流时清除粘性会话绑定
+	schedulerSnapshot *SchedulerSnapshotService
+	internal500Cache  Internal500CounterCache // INTERNAL 500 渐进惩罚计数器
 }
 
 func NewAntigravityGatewayService(
@@ -896,18 +889,16 @@ func NewAntigravityGatewayService(
 	httpUpstream HTTPUpstream,
 	settingService *SettingService,
 	internal500Cache Internal500CounterCache,
-	accountUsageService *AccountUsageService,
 ) *AntigravityGatewayService {
 	return &AntigravityGatewayService{
-		accountRepo:         accountRepo,
-		tokenProvider:       tokenProvider,
-		rateLimitService:    rateLimitService,
-		httpUpstream:        httpUpstream,
-		settingService:      settingService,
-		cache:               cache,
-		schedulerSnapshot:   schedulerSnapshot,
-		internal500Cache:    internal500Cache,
-		accountUsageService: accountUsageService,
+		accountRepo:       accountRepo,
+		tokenProvider:     tokenProvider,
+		rateLimitService:  rateLimitService,
+		httpUpstream:      httpUpstream,
+		settingService:    settingService,
+		cache:             cache,
+		schedulerSnapshot: schedulerSnapshot,
+		internal500Cache:  internal500Cache,
 	}
 }
 
@@ -1007,8 +998,14 @@ func (s *AntigravityGatewayService) getMappedModel(account *Account, requestedMo
 }
 
 // applyThinkingModelSuffix 根据 thinking 配置调整模型名
-// Sonnet 4.5 已统一映射到 4.6，此函数保留为扩展点
+// 当映射结果是 claude-sonnet-4-5 且请求开启了 thinking 时，改为 claude-sonnet-4-5-thinking
 func applyThinkingModelSuffix(mappedModel string, thinkingEnabled bool) string {
+	if !thinkingEnabled {
+		return mappedModel
+	}
+	if mappedModel == "claude-sonnet-4-5" {
+		return "claude-sonnet-4-5-thinking"
+	}
 	return mappedModel
 }
 
@@ -1744,7 +1741,7 @@ func (s *AntigravityGatewayService) Forward(ctx context.Context, c *gin.Context,
 	var clientDisconnect bool
 	if claudeReq.Stream {
 		// 客户端要求流式，直接透传转换
-		streamRes, err := s.handleClaudeStreamingResponse(c, resp, startTime, originalModel, account.ID)
+		streamRes, err := s.handleClaudeStreamingResponse(c, resp, startTime, originalModel)
 		if err != nil {
 			logger.LegacyPrintf("service.antigravity_gateway", "%s status=stream_error error=%v", prefix, err)
 			return nil, err
@@ -1754,7 +1751,7 @@ func (s *AntigravityGatewayService) Forward(ctx context.Context, c *gin.Context,
 		clientDisconnect = streamRes.clientDisconnect
 	} else {
 		// 客户端要求非流式，收集流式响应后转换返回
-		streamRes, err := s.handleClaudeStreamToNonStreaming(c, resp, startTime, originalModel, account.ID)
+		streamRes, err := s.handleClaudeStreamToNonStreaming(c, resp, startTime, originalModel)
 		if err != nil {
 			logger.LegacyPrintf("service.antigravity_gateway", "%s status=stream_collect_error error=%v", prefix, err)
 			return nil, err
@@ -3702,7 +3699,7 @@ func (s *AntigravityGatewayService) writeGoogleError(c *gin.Context, status int,
 
 // handleClaudeStreamToNonStreaming 收集上游流式响应，转换为 Claude 非流式格式返回
 // 用于处理客户端非流式请求但上游只支持流式的情况
-func (s *AntigravityGatewayService) handleClaudeStreamToNonStreaming(c *gin.Context, resp *http.Response, startTime time.Time, originalModel string, accountID int64) (*antigravityStreamResult, error) {
+func (s *AntigravityGatewayService) handleClaudeStreamToNonStreaming(c *gin.Context, resp *http.Response, startTime time.Time, originalModel string) (*antigravityStreamResult, error) {
 	scanner := bufio.NewScanner(resp.Body)
 	maxLineSize := defaultMaxLineSize
 	if s.settingService.cfg != nil && s.settingService.cfg.Gateway.MaxLineSize > 0 {
@@ -3874,7 +3871,7 @@ returnResponse:
 }
 
 // handleClaudeStreamingResponse 处理 Claude 流式响应（Gemini SSE → Claude SSE 转换）
-func (s *AntigravityGatewayService) handleClaudeStreamingResponse(c *gin.Context, resp *http.Response, startTime time.Time, originalModel string, accountID int64) (*antigravityStreamResult, error) {
+func (s *AntigravityGatewayService) handleClaudeStreamingResponse(c *gin.Context, resp *http.Response, startTime time.Time, originalModel string) (*antigravityStreamResult, error) {
 	c.Header("Content-Type", "text/event-stream")
 	c.Header("Cache-Control", "no-cache")
 	c.Header("Connection", "keep-alive")
@@ -3887,7 +3884,6 @@ func (s *AntigravityGatewayService) handleClaudeStreamingResponse(c *gin.Context
 	}
 
 	processor := antigravity.NewStreamingProcessor(originalModel)
-
 	var firstTokenMs *int
 	// 使用 Scanner 并限制单行大小，避免 ReadString 无上限导致 OOM
 	scanner := bufio.NewScanner(resp.Body)
diff --git a/backend/internal/service/antigravity_gateway_service_test.go b/backend/internal/service/antigravity_gateway_service_test.go
index 48a580ac4d..1eb1451e8a 100644
--- a/backend/internal/service/antigravity_gateway_service_test.go
+++ b/backend/internal/service/antigravity_gateway_service_test.go
@@ -926,7 +926,7 @@ func TestHandleClaudeStreamingResponse_NormalComplete(t *testing.T) {
 		fmt.Fprintln(pw, "")
 	}()
 
-	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5", 0)
+	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5")
 	_ = pr.Close()
 
 	require.NoError(t, err)
@@ -1003,7 +1003,7 @@ func TestHandleClaudeStreamingResponse_ThoughtsTokenCount(t *testing.T) {
 		fmt.Fprintln(pw, "")
 	}()
 
-	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "gemini-2.5-pro", 0)
+	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "gemini-2.5-pro")
 	_ = pr.Close()
 
 	require.NoError(t, err)
@@ -1206,7 +1206,7 @@ func TestHandleClaudeStreamingResponse_ClientDisconnect(t *testing.T) {
 		fmt.Fprintln(pw, "")
 	}()
 
-	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5", 0)
+	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5")
 	_ = pr.Close()
 
 	require.NoError(t, err)
@@ -1238,7 +1238,7 @@ func TestHandleClaudeStreamingResponse_EmptyStream(t *testing.T) {
 		fmt.Fprintln(pw, "")
 	}()
 
-	_, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5", 0)
+	_, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5")
 	_ = pr.Close()
 
 	// 应当返回 UpstreamFailoverError 而非 nil，以便上层触发 failover
@@ -1270,7 +1270,7 @@ func TestHandleClaudeStreamingResponse_ContextCanceled(t *testing.T) {
 
 	resp := &http.Response{StatusCode: http.StatusOK, Body: cancelReadCloser{}, Header: http.Header{}}
 
-	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5", 0)
+	result, err := svc.handleClaudeStreamingResponse(c, resp, time.Now(), "claude-sonnet-4-5")
 
 	require.NoError(t, err)
 	require.NotNil(t, result)
diff --git a/backend/internal/service/antigravity_model_mapping_test.go b/backend/internal/service/antigravity_model_mapping_test.go
index a151e1a139..a29000e7ed 100644
--- a/backend/internal/service/antigravity_model_mapping_test.go
+++ b/backend/internal/service/antigravity_model_mapping_test.go
@@ -69,10 +69,10 @@ func TestAntigravityGatewayService_GetMappedModel(t *testing.T) {
 			expected:       "claude-sonnet-4-6",
 		},
 		{
-			name:           "默认映射 - claude-sonnet-4-5-20250929 → claude-sonnet-4-6",
+			name:           "默认映射 - claude-sonnet-4-5-20250929 → claude-sonnet-4-5",
 			requestedModel: "claude-sonnet-4-5-20250929",
 			accountMapping: nil,
-			expected:       "claude-sonnet-4-6",
+			expected:       "claude-sonnet-4-5",
 		},
 
 		// 3. 默认映射中的透传（映射到自己）
@@ -83,10 +83,10 @@ func TestAntigravityGatewayService_GetMappedModel(t *testing.T) {
 			expected:       "claude-sonnet-4-6",
 		},
 		{
-			name:           "默认映射 - claude-sonnet-4-5 → claude-sonnet-4-6",
+			name:           "默认映射透传 - claude-sonnet-4-5",
 			requestedModel: "claude-sonnet-4-5",
 			accountMapping: nil,
-			expected:       "claude-sonnet-4-6",
+			expected:       "claude-sonnet-4-5",
 		},
 		{
 			name:           "默认映射透传 - claude-opus-4-6-thinking",
@@ -95,10 +95,10 @@ func TestAntigravityGatewayService_GetMappedModel(t *testing.T) {
 			expected:       "claude-opus-4-6-thinking",
 		},
 		{
-			name:           "默认映射 - claude-sonnet-4-5-thinking → claude-sonnet-4-6",
+			name:           "默认映射透传 - claude-sonnet-4-5-thinking",
 			requestedModel: "claude-sonnet-4-5-thinking",
 			accountMapping: nil,
-			expected:       "claude-sonnet-4-6",
+			expected:       "claude-sonnet-4-5-thinking",
 		},
 		{
 			name:           "默认映射透传 - gemini-2.5-flash",
diff --git a/backend/internal/service/antigravity_rate_limit_test.go b/backend/internal/service/antigravity_rate_limit_test.go
index d0559d7f8c..35e130dc47 100644
--- a/backend/internal/service/antigravity_rate_limit_test.go
+++ b/backend/internal/service/antigravity_rate_limit_test.go
@@ -206,7 +206,7 @@ func TestHandleUpstreamError_429_NonModelRateLimit(t *testing.T) {
 	// 但 429 兜底逻辑会使用 requestedModel 设置模型级限流
 	require.Nil(t, result)
 	require.Len(t, repo.modelRateLimitCalls, 1)
-	require.Equal(t, "claude-sonnet-4-6", repo.modelRateLimitCalls[0].modelKey)
+	require.Equal(t, "claude-sonnet-4-5", repo.modelRateLimitCalls[0].modelKey)
 }
 
 // TestHandleUpstreamError_429_NonModelRateLimit_UsesMappedModelKey 测试 429 非模型限流场景
@@ -832,7 +832,7 @@ func TestAntigravityRetryLoop_PreCheck_SwitchesWhenRateLimited(t *testing.T) {
 		Concurrency: 1,
 		Extra: map[string]any{
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limit_reset_at": time.Now().Add(2 * time.Second).Format(time.RFC3339),
 				},
 			},
@@ -875,7 +875,7 @@ func TestAntigravityRetryLoop_PreCheck_SwitchesWhenRemainingLong(t *testing.T) {
 		Concurrency: 1,
 		Extra: map[string]any{
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limit_reset_at": time.Now().Add(11 * time.Second).Format(time.RFC3339),
 				},
 			},
diff --git a/backend/internal/service/antigravity_single_account_retry_test.go b/backend/internal/service/antigravity_single_account_retry_test.go
index 6ac19b610a..675e9c0cbf 100644
--- a/backend/internal/service/antigravity_single_account_retry_test.go
+++ b/backend/internal/service/antigravity_single_account_retry_test.go
@@ -689,7 +689,7 @@ func TestAntigravityRetryLoop_PreCheck_SingleAccountRetry_SkipsRateLimit(t *test
 		Concurrency: 1,
 		Extra: map[string]any{
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limit_reset_at": time.Now().Add(30 * time.Second).Format(time.RFC3339),
 				},
 			},
@@ -733,7 +733,7 @@ func TestAntigravityRetryLoop_PreCheck_NoSingleAccountRetry_SwitchesOnRateLimit(
 		Concurrency: 1,
 		Extra: map[string]any{
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5": map[string]any{
 					"rate_limit_reset_at": time.Now().Add(30 * time.Second).Format(time.RFC3339),
 				},
 			},
diff --git a/backend/internal/service/antigravity_thinking_test.go b/backend/internal/service/antigravity_thinking_test.go
index 34b3bf9219..b3952ee4a8 100644
--- a/backend/internal/service/antigravity_thinking_test.go
+++ b/backend/internal/service/antigravity_thinking_test.go
@@ -27,12 +27,12 @@ func TestApplyThinkingModelSuffix(t *testing.T) {
 			expected:        "claude-opus-4-6-thinking",
 		},
 
-		// Thinking 开启 + claude-sonnet-4-5：Sonnet 4.5 已统一映射到 4.6，不再特殊处理
+		// Thinking 开启 + claude-sonnet-4-5：自动添加后缀
 		{
-			name:            "thinking enabled - claude-sonnet-4-5 unchanged (mapped to 4.6 elsewhere)",
+			name:            "thinking enabled - claude-sonnet-4-5 becomes thinking version",
 			mappedModel:     "claude-sonnet-4-5",
 			thinkingEnabled: true,
-			expected:        "claude-sonnet-4-5",
+			expected:        "claude-sonnet-4-5-thinking",
 		},
 
 		// Thinking 开启 + 其他模型：保持原样
diff --git a/backend/internal/service/gateway_service_antigravity_whitelist_test.go b/backend/internal/service/gateway_service_antigravity_whitelist_test.go
index 98b0680119..c078be3264 100644
--- a/backend/internal/service/gateway_service_antigravity_whitelist_test.go
+++ b/backend/internal/service/gateway_service_antigravity_whitelist_test.go
@@ -103,8 +103,7 @@ func TestGatewayService_isModelSupportedByAccountWithContext_ThinkingMode(t *tes
 			expected:        false,
 		},
 		// 场景 3: 配置 claude-sonnet-4-5（非 thinking），请求 claude-sonnet-4-5 + thinking=true
-		// applyThinkingModelSuffix 已简化为直通，不再添加 -thinking 后缀
-		// 最终模型名仍为 claude-sonnet-4-5，在 mapping 中 → 返回 true
+		// 最终模型名 = claude-sonnet-4-5-thinking，不在 mapping 中，应该不匹配
 		{
 			name: "thinking_enabled_no_match_non_thinking_mapping",
 			modelMapping: map[string]any{
@@ -112,7 +111,7 @@ func TestGatewayService_isModelSupportedByAccountWithContext_ThinkingMode(t *tes
 			},
 			requestedModel:  "claude-sonnet-4-5",
 			thinkingEnabled: true,
-			expected:        true,
+			expected:        false,
 		},
 		// 场景 4: 配置两种模型，请求 claude-sonnet-4-5 + thinking=true，应该匹配 thinking 版本
 		{
diff --git a/backend/internal/service/model_rate_limit_test.go b/backend/internal/service/model_rate_limit_test.go
index 7af3f6c788..b79b9688ac 100644
--- a/backend/internal/service/model_rate_limit_test.go
+++ b/backend/internal/service/model_rate_limit_test.go
@@ -181,12 +181,11 @@ func TestIsModelRateLimited_Antigravity_ThinkingAffectsModelKey(t *testing.T) {
 	now := time.Now()
 	future := now.Add(10 * time.Minute).Format(time.RFC3339)
 
-	// Sonnet 4.5 now maps to Sonnet 4.6, so rate limit key is claude-sonnet-4-6
 	account := &Account{
 		Platform: PlatformAntigravity,
 		Extra: map[string]any{
 			modelRateLimitsKey: map[string]any{
-				"claude-sonnet-4-6": map[string]any{
+				"claude-sonnet-4-5-thinking": map[string]any{
 					"rate_limit_reset_at": future,
 				},
 			},
@@ -344,7 +343,7 @@ func TestGetRateLimitRemainingTime(t *testing.T) {
 				Platform: PlatformAntigravity,
 				Extra: map[string]any{
 					modelRateLimitsKey: map[string]any{
-						"claude-sonnet-4-6": map[string]any{
+						"claude-sonnet-4-5": map[string]any{
 							"rate_limit_reset_at": future15m,
 						},
 					},
@@ -360,7 +359,7 @@ func TestGetRateLimitRemainingTime(t *testing.T) {
 				Platform: PlatformAntigravity,
 				Extra: map[string]any{
 					modelRateLimitsKey: map[string]any{
-						"claude-sonnet-4-6": map[string]any{
+						"claude-sonnet-4-5": map[string]any{
 							"rate_limit_reset_at": future5m,
 						},
 					},
diff --git a/backend/migrations/056_add_sonnet46_to_model_mapping.sql b/backend/migrations/056_add_sonnet46_to_model_mapping.sql
deleted file mode 100644
index aa7657d71a..0000000000
--- a/backend/migrations/056_add_sonnet46_to_model_mapping.sql
+++ /dev/null
@@ -1,42 +0,0 @@
--- Add claude-sonnet-4-6 to model_mapping for all Antigravity accounts
---
--- Background:
--- Antigravity now supports claude-sonnet-4-6
---
--- Strategy:
--- Directly overwrite the entire model_mapping with updated mappings
--- This ensures consistency with DefaultAntigravityModelMapping in constants.go
-
-UPDATE accounts
-SET credentials = jsonb_set(
-    credentials,
-    '{model_mapping}',
-    '{
-        "claude-opus-4-6-thinking": "claude-opus-4-6-thinking",
-        "claude-opus-4-6": "claude-opus-4-6-thinking",
-        "claude-opus-4-5-thinking": "claude-opus-4-6-thinking",
-        "claude-opus-4-5-20251101": "claude-opus-4-6-thinking",
-        "claude-sonnet-4-6": "claude-sonnet-4-6",
-        "claude-sonnet-4-5": "claude-sonnet-4-5",
-        "claude-sonnet-4-5-thinking": "claude-sonnet-4-5-thinking",
-        "claude-sonnet-4-5-20250929": "claude-sonnet-4-5",
-        "claude-haiku-4-5": "claude-sonnet-4-5",
-        "claude-haiku-4-5-20251001": "claude-sonnet-4-5",
-        "gemini-2.5-flash": "gemini-2.5-flash",
-        "gemini-2.5-flash-lite": "gemini-2.5-flash-lite",
-        "gemini-2.5-flash-thinking": "gemini-2.5-flash-thinking",
-        "gemini-2.5-pro": "gemini-2.5-pro",
-        "gemini-3-flash": "gemini-3-flash",
-        "gemini-3-pro-high": "gemini-3-pro-high",
-        "gemini-3-pro-low": "gemini-3-pro-low",
-        "gemini-3-pro-image": "gemini-3-pro-image",
-        "gemini-3-flash-preview": "gemini-3-flash",
-        "gemini-3-pro-preview": "gemini-3-pro-high",
-        "gemini-3-pro-image-preview": "gemini-3-pro-image",
-        "gpt-oss-120b-medium": "gpt-oss-120b-medium",
-        "tab_flash_lite_preview": "tab_flash_lite_preview"
-    }'::jsonb
-)
-WHERE platform = 'antigravity'
-  AND deleted_at IS NULL
-  AND credentials->'model_mapping' IS NOT NULL;
diff --git a/backend/migrations/057_add_gemini31_pro_to_model_mapping.sql b/backend/migrations/057_add_gemini31_pro_to_model_mapping.sql
deleted file mode 100644
index 6305e717bc..0000000000
--- a/backend/migrations/057_add_gemini31_pro_to_model_mapping.sql
+++ /dev/null
@@ -1,45 +0,0 @@
--- Add gemini-3.1-pro-high, gemini-3.1-pro-low, gemini-3.1-pro-preview to model_mapping
---
--- Background:
--- Antigravity now supports gemini-3.1-pro-high and gemini-3.1-pro-low
---
--- Strategy:
--- Directly overwrite the entire model_mapping with updated mappings
--- This ensures consistency with DefaultAntigravityModelMapping in constants.go
-
-UPDATE accounts
-SET credentials = jsonb_set(
-    credentials,
-    '{model_mapping}',
-    '{
-        "claude-opus-4-6-thinking": "claude-opus-4-6-thinking",
-        "claude-opus-4-6": "claude-opus-4-6-thinking",
-        "claude-opus-4-5-thinking": "claude-opus-4-6-thinking",
-        "claude-opus-4-5-20251101": "claude-opus-4-6-thinking",
-        "claude-sonnet-4-6": "claude-sonnet-4-6",
-        "claude-sonnet-4-5": "claude-sonnet-4-5",
-        "claude-sonnet-4-5-thinking": "claude-sonnet-4-5-thinking",
-        "claude-sonnet-4-5-20250929": "claude-sonnet-4-5",
-        "claude-haiku-4-5": "claude-sonnet-4-5",
-        "claude-haiku-4-5-20251001": "claude-sonnet-4-5",
-        "gemini-2.5-flash": "gemini-2.5-flash",
-        "gemini-2.5-flash-lite": "gemini-2.5-flash-lite",
-        "gemini-2.5-flash-thinking": "gemini-2.5-flash-thinking",
-        "gemini-2.5-pro": "gemini-2.5-pro",
-        "gemini-3-flash": "gemini-3-flash",
-        "gemini-3-pro-high": "gemini-3-pro-high",
-        "gemini-3-pro-low": "gemini-3-pro-low",
-        "gemini-3-pro-image": "gemini-3-pro-image",
-        "gemini-3-flash-preview": "gemini-3-flash",
-        "gemini-3-pro-preview": "gemini-3-pro-high",
-        "gemini-3-pro-image-preview": "gemini-3-pro-image",
-        "gemini-3.1-pro-high": "gemini-3.1-pro-high",
-        "gemini-3.1-pro-low": "gemini-3.1-pro-low",
-        "gemini-3.1-pro-preview": "gemini-3.1-pro-high",
-        "gpt-oss-120b-medium": "gpt-oss-120b-medium",
-        "tab_flash_lite_preview": "tab_flash_lite_preview"
-    }'::jsonb
-)
-WHERE platform = 'antigravity'
-  AND deleted_at IS NULL
-  AND credentials->'model_mapping' IS NOT NULL;
diff --git a/backend/migrations/060_add_group_simulate_claude_max.sql b/backend/migrations/060_add_group_simulate_claude_max.sql
deleted file mode 100644
index 55662dfdac..0000000000
--- a/backend/migrations/060_add_group_simulate_claude_max.sql
+++ /dev/null
@@ -1,3 +0,0 @@
-ALTER TABLE groups
-    ADD COLUMN IF NOT EXISTS simulate_claude_max_enabled BOOLEAN NOT NULL DEFAULT FALSE;
-
diff --git a/backend/migrations/070_add_sora_tasks.sql b/backend/migrations/070_add_sora_tasks.sql
deleted file mode 100644
index 3a484976d9..0000000000
--- a/backend/migrations/070_add_sora_tasks.sql
+++ /dev/null
@@ -1,36 +0,0 @@
--- +migrate Up
-
--- Sora 异步任务表：记录视频/角色/图片生成任务状态
-CREATE TABLE IF NOT EXISTS sora_tasks (
-    id               VARCHAR(64) PRIMARY KEY,                   -- 对外 ID (video_xxx / char_xxx / img_xxx)
-    account_id       BIGINT NOT NULL,                           -- 使用的账号 ID
-    api_key_id       BIGINT,                                    -- 调用方 API Key ID
-    upstream_task_id VARCHAR(128) NOT NULL DEFAULT '',           -- 上游任务 ID
-    object_type      VARCHAR(16) NOT NULL DEFAULT 'video',      -- video / character / image
-    model            VARCHAR(64) NOT NULL,                      -- 请求模型名
-    prompt           TEXT NOT NULL DEFAULT '',                   -- 生成提示词
-    status           VARCHAR(16) NOT NULL DEFAULT 'queued',     -- queued / in_progress / completed / failed
-    progress         INT NOT NULL DEFAULT 0,                    -- 进度 0-100
-    video_url        TEXT NOT NULL DEFAULT '',                   -- 原始上游下载 URL
-    stored_key       TEXT NOT NULL DEFAULT '',                   -- 存储后的 key（本地相对路径或 S3 object key）
-    storage_type     VARCHAR(16) NOT NULL DEFAULT '',            -- 存储类型：local / s3 / gdrive / 空=未存储
-    share_id         VARCHAR(128) NOT NULL DEFAULT '',           -- 可分享 ID
-    character_info   JSONB,                                     -- 角色信息 {username, display_name}
-    error_message    TEXT NOT NULL DEFAULT '',                   -- 错误信息
-    error_type       VARCHAR(64) NOT NULL DEFAULT '',            -- 错误类型
-    request_body     JSONB,                                     -- 原始请求体（用于上游重放）
-    seconds          VARCHAR(8) NOT NULL DEFAULT '',             -- 视频时长
-    size             VARCHAR(16) NOT NULL DEFAULT '',            -- 分辨率 (如 1920x1080)
-    created_at       TIMESTAMPTZ NOT NULL DEFAULT NOW(),
-    completed_at     TIMESTAMPTZ
-);
-
--- 仅索引未完成任务，供后台 worker 轮询
-CREATE INDEX IF NOT EXISTS idx_sora_tasks_pending ON sora_tasks(status) WHERE status IN ('queued', 'in_progress');
-
--- 按 API Key 查询
-CREATE INDEX IF NOT EXISTS idx_sora_tasks_api_key ON sora_tasks(api_key_id) WHERE api_key_id IS NOT NULL;
-
--- +migrate Down
-
-DROP TABLE IF EXISTS sora_tasks;
diff --git a/backend/migrations/077_rollback_sonnet_to_original.sql b/backend/migrations/077_rollback_sonnet_to_original.sql
deleted file mode 100644
index 13316e0b72..0000000000
--- a/backend/migrations/077_rollback_sonnet_to_original.sql
+++ /dev/null
@@ -1,18 +0,0 @@
--- 077: Rollback Sonnet/Haiku mappings from Opus back to original values
--- Reverts migration 077_migrate_sonnet_to_opus46_thinking.sql
--- Restores: Sonnet→Sonnet, Haiku→Sonnet (original behavior)
-
-UPDATE accounts
-SET credentials = jsonb_set(
-    credentials, '{model_mapping}',
-    (credentials->'model_mapping')
-        || '{"claude-sonnet-4-6": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-sonnet-4-5": "claude-sonnet-4-5"}'::jsonb
-        || '{"claude-sonnet-4-5-thinking": "claude-sonnet-4-5-thinking"}'::jsonb
-        || '{"claude-sonnet-4-5-20250929": "claude-sonnet-4-5"}'::jsonb
-        || '{"claude-haiku-4-5": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-haiku-4-5-20251001": "claude-sonnet-4-6"}'::jsonb
-)
-WHERE platform = 'antigravity'
-  AND deleted_at IS NULL
-  AND credentials->'model_mapping' IS NOT NULL;
diff --git a/backend/migrations/078_migrate_sonnet45_to_sonnet46.sql b/backend/migrations/078_migrate_sonnet45_to_sonnet46.sql
deleted file mode 100644
index d617ca4a16..0000000000
--- a/backend/migrations/078_migrate_sonnet45_to_sonnet46.sql
+++ /dev/null
@@ -1,16 +0,0 @@
--- 078: Migrate Sonnet 4.5 → Sonnet 4.6 for all antigravity accounts
--- Sonnet 4.5 is no longer reliable on upstream, redirect to 4.6
-
-UPDATE accounts
-SET credentials = jsonb_set(
-    credentials, '{model_mapping}',
-    (credentials->'model_mapping')
-        || '{"claude-sonnet-4-5": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-sonnet-4-5-thinking": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-sonnet-4-5-20250929": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-haiku-4-5": "claude-sonnet-4-6"}'::jsonb
-        || '{"claude-haiku-4-5-20251001": "claude-sonnet-4-6"}'::jsonb
-)
-WHERE platform = 'antigravity'
-  AND deleted_at IS NULL
-  AND credentials->'model_mapping' IS NOT NULL;
diff --git a/frontend/src/composables/useModelWhitelist.ts b/frontend/src/composables/useModelWhitelist.ts
index 9594313914..d16ed97707 100644
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -283,10 +283,10 @@ const geminiPresetMappings = [
 // Antigravity 预设映射（支持通配符）
 const antigravityPresetMappings = [
   // Claude 通配符映射
-  { label: 'Claude→Sonnet4.6', from: 'claude-*', to: 'claude-sonnet-4-6', color: 'bg-blue-100 text-blue-700 hover:bg-blue-200 dark:bg-blue-900/30 dark:text-blue-400' },
-  { label: 'Sonnet→4.6', from: 'claude-sonnet-*', to: 'claude-sonnet-4-6', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
+  { label: 'Claude→Sonnet', from: 'claude-*', to: 'claude-sonnet-4-5', color: 'bg-blue-100 text-blue-700 hover:bg-blue-200 dark:bg-blue-900/30 dark:text-blue-400' },
+  { label: 'Sonnet→Sonnet', from: 'claude-sonnet-*', to: 'claude-sonnet-4-5', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
   { label: 'Opus→Opus', from: 'claude-opus-*', to: 'claude-opus-4-6-thinking', color: 'bg-purple-100 text-purple-700 hover:bg-purple-200 dark:bg-purple-900/30 dark:text-purple-400' },
-  { label: 'Haiku→Sonnet4.6', from: 'claude-haiku-*', to: 'claude-sonnet-4-6', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' },
+  { label: 'Haiku→Sonnet', from: 'claude-haiku-*', to: 'claude-sonnet-4-5', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' },
   { label: 'Sonnet4→4.6', from: 'claude-sonnet-4-20250514', to: 'claude-sonnet-4-6', color: 'bg-sky-100 text-sky-700 hover:bg-sky-200 dark:bg-sky-900/30 dark:text-sky-400' },
   { label: 'Sonnet4.5→4.6', from: 'claude-sonnet-4-5-20250929', to: 'claude-sonnet-4-6', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' },
   { label: 'Sonnet3.5→4.6', from: 'claude-3-5-sonnet-20241022', to: 'claude-sonnet-4-6', color: 'bg-teal-100 text-teal-700 hover:bg-teal-200 dark:bg-teal-900/30 dark:text-teal-400' },
@@ -307,7 +307,7 @@ const antigravityPresetMappings = [
   { label: '2.5-Flash-Lite透传', from: 'gemini-2.5-flash-lite', to: 'gemini-2.5-flash-lite', color: 'bg-green-100 text-green-700 hover:bg-green-200 dark:bg-green-900/30 dark:text-green-400' },
   // 精确映射
   { label: 'Sonnet 4.6', from: 'claude-sonnet-4-6', to: 'claude-sonnet-4-6', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' },
-  { label: 'Sonnet 4.5→4.6', from: 'claude-sonnet-4-5', to: 'claude-sonnet-4-6', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' },
+  { label: 'Sonnet 4.5', from: 'claude-sonnet-4-5', to: 'claude-sonnet-4-5', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' },
   { label: 'Opus 4.6', from: 'claude-opus-4-6', to: 'claude-opus-4-6-thinking', color: 'bg-pink-100 text-pink-700 hover:bg-pink-200 dark:bg-pink-900/30 dark:text-pink-400' },
   { label: 'Opus 4.6-thinking', from: 'claude-opus-4-6-thinking', to: 'claude-opus-4-6-thinking', color: 'bg-pink-100 text-pink-700 hover:bg-pink-200 dark:bg-pink-900/30 dark:text-pink-400' }
 ]