feat: image output token billing, channel-mapped billing source, credits balance precheck

- Parse candidatesTokensDetails from Gemini API to separate image/text output tokens
- Add image_output_tokens and image_output_cost to usage_log (migration 089)
- Support per-image-token pricing via output_cost_per_image_token from model pricing data
- Channel pricing ImageOutputPrice override works in token billing mode
- Auto-fill image_output_price in channel pricing form from model defaults
- Add "channel_mapped" billing model source as new default (migration 088)
- Bills by model name after channel mapping, before account mapping
- Fix channel cache error TTL sign error (115s → 5s)
- Fix Update channel only invalidating new groups, not removed groups
- Fix frontend model_mapping clearing sending undefined instead of {}
- Credits balance precheck via shared AccountUsageService cache before injection
- Skip credits injection for accounts with insufficient balance
- Don't mark credits exhausted for "exhausted your capacity on this model" 429s

backend/cmd/server/wire_gen.go

@@ -139,11 +139,11 @@ 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)
 	tlsFingerprintProfileRepository := repository.NewTLSFingerprintProfileRepository(client)
 	tlsFingerprintProfileCache := repository.NewTLSFingerprintProfileCache(redisClient)
 	tlsFingerprintProfileService := service.NewTLSFingerprintProfileService(tlsFingerprintProfileRepository, tlsFingerprintProfileCache)
 	accountUsageService := service.NewAccountUsageService(accountRepository, usageLogRepository, claudeUsageFetcher, geminiQuotaService, antigravityQuotaFetcher, usageCache, identityCache, tlsFingerprintProfileService)
+	antigravityGatewayService := service.NewAntigravityGatewayService(accountRepository, gatewayCache, schedulerSnapshotService, antigravityTokenProvider, rateLimitService, httpUpstream, settingService, internal500CounterCache, accountUsageService)
 	accountTestService := service.NewAccountTestService(accountRepository, geminiTokenProvider, antigravityGatewayService, httpUpstream, configConfig, tlsFingerprintProfileService)
 	crsSyncService := service.NewCRSSyncService(accountRepository, proxyRepository, oAuthService, openAIOAuthService, geminiOAuthService, configConfig)
 	sessionLimitCache := repository.ProvideSessionLimitCache(redisClient, configConfig)

backend/internal/handler/admin/channel_handler.go

@@ -31,7 +31,7 @@ type createChannelRequest struct {
 	GroupIDs           []int64                      `json:"group_ids"`
 	ModelPricing       []channelModelPricingRequest `json:"model_pricing"`
 	ModelMapping       map[string]map[string]string `json:"model_mapping"`
-	BillingModelSource string                       `json:"billing_model_source" binding:"omitempty,oneof=requested upstream"`
+	BillingModelSource string                       `json:"billing_model_source" binding:"omitempty,oneof=requested upstream channel_mapped"`
 	RestrictModels     bool                         `json:"restrict_models"`
 }
 
@@ -42,7 +42,7 @@ type updateChannelRequest struct {
 	GroupIDs           *[]int64                      `json:"group_ids"`
 	ModelPricing       *[]channelModelPricingRequest `json:"model_pricing"`
 	ModelMapping       map[string]map[string]string  `json:"model_mapping"`
-	BillingModelSource string                        `json:"billing_model_source" binding:"omitempty,oneof=requested upstream"`
+	BillingModelSource string                        `json:"billing_model_source" binding:"omitempty,oneof=requested upstream channel_mapped"`
 	RestrictModels     *bool                         `json:"restrict_models"`
 }
 
@@ -129,7 +129,7 @@ func channelToResponse(ch *service.Channel) *channelResponse {
 	}
 	resp.BillingModelSource = ch.BillingModelSource
 	if resp.BillingModelSource == "" {
-		resp.BillingModelSource = "requested"
+		resp.BillingModelSource = "channel_mapped"
 	}
 	if resp.GroupIDs == nil {
 		resp.GroupIDs = []int64{}
@@ -393,5 +393,6 @@ func (h *ChannelHandler) GetModelDefaultPricing(c *gin.Context) {
 		"output_price":       pricing.OutputPricePerToken,
 		"cache_write_price":  pricing.CacheCreationPricePerToken,
 		"cache_read_price":   pricing.CacheReadPricePerToken,
+		"image_output_price": pricing.ImageOutputPricePerToken,
 	})
 }

backend/internal/handler/admin/channel_handler_test.go

@@ -106,7 +106,7 @@ func TestChannelToResponse_EmptyDefaults(t *testing.T) {
 	}
 
 	resp := channelToResponse(ch)
-	require.Equal(t, "requested", resp.BillingModelSource)
+	require.Equal(t, "channel_mapped", resp.BillingModelSource)
 	require.NotNil(t, resp.GroupIDs)
 	require.Empty(t, resp.GroupIDs)
 	require.NotNil(t, resp.ModelMapping)

backend/internal/pkg/antigravity/claude_types.go

@@ -125,6 +125,7 @@ type ClaudeUsage struct {
 	OutputTokens             int `json:"output_tokens"`
 	CacheCreationInputTokens int `json:"cache_creation_input_tokens,omitempty"`
 	CacheReadInputTokens     int `json:"cache_read_input_tokens,omitempty"`
+	ImageOutputTokens        int `json:"image_output_tokens,omitempty"`
 }
 
 // ClaudeError Claude 错误响应

backend/internal/pkg/antigravity/gemini_types.go

@@ -149,6 +149,12 @@ type GeminiCandidate struct {
 	GroundingMetadata *GeminiGroundingMetadata `json:"groundingMetadata,omitempty"`
 }
 
+// GeminiTokenDetail Gemini token 详情（按模态分类）
+type GeminiTokenDetail struct {
+	Modality   string `json:"modality"`
+	TokenCount int    `json:"tokenCount"`
+}
+
 // GeminiUsageMetadata Gemini 用量元数据
 type GeminiUsageMetadata struct {
 	PromptTokenCount        int                 `json:"promptTokenCount,omitempty"`
@@ -156,6 +162,18 @@ type GeminiUsageMetadata struct {
 	CachedContentTokenCount int                 `json:"cachedContentTokenCount,omitempty"`
 	TotalTokenCount         int                 `json:"totalTokenCount,omitempty"`
 	ThoughtsTokenCount      int                 `json:"thoughtsTokenCount,omitempty"` // thinking tokens（按输出价格计费）
+	CandidatesTokensDetails []GeminiTokenDetail `json:"candidatesTokensDetails,omitempty"`
+	PromptTokensDetails     []GeminiTokenDetail `json:"promptTokensDetails,omitempty"`
+}
+
+// ImageOutputTokens 从 CandidatesTokensDetails 中提取 IMAGE 模态的 token 数
+func (m *GeminiUsageMetadata) ImageOutputTokens() int {
+	for _, d := range m.CandidatesTokensDetails {
+		if d.Modality == "IMAGE" {
+			return d.TokenCount
+		}
+	}
+	return 0
 }
 
 // GeminiGroundingMetadata Gemini grounding 元数据（Web Search）

backend/internal/pkg/antigravity/response_transformer.go

@@ -284,6 +284,7 @@ func (p *NonStreamingProcessor) buildResponse(geminiResp *GeminiResponse, respon
 		usage.InputTokens = geminiResp.UsageMetadata.PromptTokenCount - cached
 		usage.OutputTokens = geminiResp.UsageMetadata.CandidatesTokenCount + geminiResp.UsageMetadata.ThoughtsTokenCount
 		usage.CacheReadInputTokens = cached
+		usage.ImageOutputTokens = geminiResp.UsageMetadata.ImageOutputTokens()
 	}
 
 	// 生成响应 ID

backend/internal/pkg/antigravity/stream_transformer.go

@@ -35,6 +35,7 @@ type StreamingProcessor struct {
 	inputTokens       int
 	outputTokens      int
 	cacheReadTokens   int
+	imageOutputTokens int
 }
 
 // NewStreamingProcessor 创建流式响应处理器
@@ -45,6 +46,28 @@ func NewStreamingProcessor(originalModel string) *StreamingProcessor {
 	}
 }
 
+// SetUsageMapHook sets an optional hook that modifies usage maps before they are emitted.
+func (p *StreamingProcessor) SetUsageMapHook(fn UsageMapHook) {
+	p.usageMapHook = fn
+}
+
+func usageToMap(u ClaudeUsage) map[string]any {
+	m := map[string]any{
+		"input_tokens":  u.InputTokens,
+		"output_tokens": u.OutputTokens,
+	}
+	if u.CacheCreationInputTokens > 0 {
+		m["cache_creation_input_tokens"] = u.CacheCreationInputTokens
+	}
+	if u.CacheReadInputTokens > 0 {
+		m["cache_read_input_tokens"] = u.CacheReadInputTokens
+	}
+	if u.ImageOutputTokens > 0 {
+		m["image_output_tokens"] = u.ImageOutputTokens
+	}
+	return m
+}
+
 // ProcessLine 处理 SSE 行，返回 Claude SSE 事件
 func (p *StreamingProcessor) ProcessLine(line string) []byte {
 	line = strings.TrimSpace(line)
@@ -87,6 +110,7 @@ func (p *StreamingProcessor) ProcessLine(line string) []byte {
 		p.inputTokens = geminiResp.UsageMetadata.PromptTokenCount - cached
 		p.outputTokens = geminiResp.UsageMetadata.CandidatesTokenCount + geminiResp.UsageMetadata.ThoughtsTokenCount
 		p.cacheReadTokens = cached
+		p.imageOutputTokens = geminiResp.UsageMetadata.ImageOutputTokens()
 	}
 
 	// 处理 parts
@@ -127,6 +151,7 @@ func (p *StreamingProcessor) Finish() ([]byte, *ClaudeUsage) {
 		InputTokens:          p.inputTokens,
 		OutputTokens:         p.outputTokens,
 		CacheReadInputTokens: p.cacheReadTokens,
+		ImageOutputTokens:    p.imageOutputTokens,
 	}
 
 	if !p.messageStartSent {
@@ -158,6 +183,7 @@ func (p *StreamingProcessor) emitMessageStart(v1Resp *V1InternalResponse) []byte
 		usage.InputTokens = v1Resp.Response.UsageMetadata.PromptTokenCount - cached
 		usage.OutputTokens = v1Resp.Response.UsageMetadata.CandidatesTokenCount + v1Resp.Response.UsageMetadata.ThoughtsTokenCount
 		usage.CacheReadInputTokens = cached
+		usage.ImageOutputTokens = v1Resp.Response.UsageMetadata.ImageOutputTokens()
 	}
 
 	responseID := v1Resp.ResponseID
@@ -485,6 +511,7 @@ func (p *StreamingProcessor) emitFinish(finishReason string) []byte {
 		InputTokens:          p.inputTokens,
 		OutputTokens:         p.outputTokens,
 		CacheReadInputTokens: p.cacheReadTokens,
+		ImageOutputTokens:    p.imageOutputTokens,
 	}
 
 	deltaEvent := map[string]any{

backend/internal/repository/usage_log_repo.go

@@ -28,7 +28,7 @@ import (
 	gocache "github.com/patrickmn/go-cache"
 )
 
-const usageLogSelectColumns = "id, user_id, api_key_id, account_id, request_id, model, requested_model, upstream_model, group_id, subscription_id, input_tokens, output_tokens, cache_creation_tokens, cache_read_tokens, cache_creation_5m_tokens, cache_creation_1h_tokens, input_cost, output_cost, cache_creation_cost, cache_read_cost, total_cost, actual_cost, rate_multiplier, account_rate_multiplier, billing_type, request_type, stream, openai_ws_mode, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, media_type, service_tier, reasoning_effort, inbound_endpoint, upstream_endpoint, cache_ttl_overridden, channel_id, model_mapping_chain, billing_tier, billing_mode, created_at"
+const usageLogSelectColumns = "id, user_id, api_key_id, account_id, request_id, model, requested_model, upstream_model, group_id, subscription_id, input_tokens, output_tokens, cache_creation_tokens, cache_read_tokens, cache_creation_5m_tokens, cache_creation_1h_tokens, image_output_tokens, image_output_cost, input_cost, output_cost, cache_creation_cost, cache_read_cost, total_cost, actual_cost, rate_multiplier, account_rate_multiplier, billing_type, request_type, stream, openai_ws_mode, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, media_type, service_tier, reasoning_effort, inbound_endpoint, upstream_endpoint, cache_ttl_overridden, channel_id, model_mapping_chain, billing_tier, billing_mode, created_at"
 
 // usageLogInsertArgTypes must stay in the same order as:
 //  1. prepareUsageLogInsert().args
@@ -53,6 +53,8 @@ var usageLogInsertArgTypes = [...]string{
 	"integer",     // cache_read_tokens
 	"integer",     // cache_creation_5m_tokens
 	"integer",     // cache_creation_1h_tokens
+	"integer",     // image_output_tokens
+	"numeric",     // image_output_cost
 	"numeric",     // input_cost
 	"numeric",     // output_cost
 	"numeric",     // cache_creation_cost
@@ -330,6 +332,8 @@ func (r *usageLogRepository) createSingle(ctx context.Context, sqlq sqlExecutor,
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -363,9 +367,9 @@ func (r *usageLogRepository) createSingle(ctx context.Context, sqlq sqlExecutor,
 			$1, $2, $3, $4, $5, $6, $7,
 			$8, $9,
 			$10, $11, $12, $13,
-			$14, $15,
-			$16, $17, $18, $19, $20, $21,
-			$22, $23, $24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38, $39, $40, $41, $42, $43, $44
+			$14, $15, $16, $17,
+			$18, $19, $20, $21, $22, $23,
+			$24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38, $39, $40, $41, $42, $43, $44, $45, $46
 		)
 		ON CONFLICT (request_id, api_key_id) DO NOTHING
 		RETURNING id, created_at
@@ -766,6 +770,8 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -797,7 +803,7 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 			created_at
 		) AS (VALUES `)
 
-	args := make([]any, 0, len(keys)*45)
+	args := make([]any, 0, len(keys)*47)
 	argPos := 1
 	for idx, key := range keys {
 		if idx > 0 {
@@ -841,6 +847,8 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 				cache_read_tokens,
 				cache_creation_5m_tokens,
 				cache_creation_1h_tokens,
+				image_output_tokens,
+				image_output_cost,
 				input_cost,
 				output_cost,
 				cache_creation_cost,
@@ -887,6 +895,8 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 				cache_read_tokens,
 				cache_creation_5m_tokens,
 				cache_creation_1h_tokens,
+				image_output_tokens,
+				image_output_cost,
 				input_cost,
 				output_cost,
 				cache_creation_cost,
@@ -973,6 +983,8 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -1004,7 +1016,7 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			created_at
 		) AS (VALUES `)
 
-	args := make([]any, 0, len(preparedList)*44)
+	args := make([]any, 0, len(preparedList)*46)
 	argPos := 1
 	for idx, prepared := range preparedList {
 		if idx > 0 {
@@ -1045,6 +1057,8 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -1091,6 +1105,8 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -1145,6 +1161,8 @@ func execUsageLogInsertNoResult(ctx context.Context, sqlq sqlExecutor, prepared
 			cache_read_tokens,
 			cache_creation_5m_tokens,
 			cache_creation_1h_tokens,
+			image_output_tokens,
+			image_output_cost,
 			input_cost,
 			output_cost,
 			cache_creation_cost,
@@ -1178,9 +1196,9 @@ func execUsageLogInsertNoResult(ctx context.Context, sqlq sqlExecutor, prepared
 			$1, $2, $3, $4, $5, $6, $7,
 			$8, $9,
 			$10, $11, $12, $13,
-			$14, $15,
-			$16, $17, $18, $19, $20, $21,
-			$22, $23, $24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38, $39, $40, $41, $42, $43, $44
+			$14, $15, $16, $17,
+			$18, $19, $20, $21, $22, $23,
+			$24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38, $39, $40, $41, $42, $43, $44, $45, $46
 		)
 		ON CONFLICT (request_id, api_key_id) DO NOTHING
 	`, prepared.args...)
@@ -1248,6 +1266,8 @@ func prepareUsageLogInsert(log *service.UsageLog) usageLogInsertPrepared {
 			log.CacheReadTokens,
 			log.CacheCreation5mTokens,
 			log.CacheCreation1hTokens,
+			log.ImageOutputTokens,
+			log.ImageOutputCost,
 			log.InputCost,
 			log.OutputCost,
 			log.CacheCreationCost,
@@ -4011,6 +4031,8 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		cacheReadTokens       int
 		cacheCreation5m       int
 		cacheCreation1h       int
+		imageOutputTokens     int
+		imageOutputCost       float64
 		inputCost             float64
 		outputCost            float64
 		cacheCreationCost     float64
@@ -4059,6 +4081,8 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		&cacheReadTokens,
 		&cacheCreation5m,
 		&cacheCreation1h,
+		&imageOutputTokens,
+		&imageOutputCost,
 		&inputCost,
 		&outputCost,
 		&cacheCreationCost,
@@ -4105,6 +4129,8 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		CacheReadTokens:       cacheReadTokens,
 		CacheCreation5mTokens: cacheCreation5m,
 		CacheCreation1hTokens: cacheCreation1h,
+		ImageOutputTokens:     imageOutputTokens,
+		ImageOutputCost:       imageOutputCost,
 		InputCost:             inputCost,
 		OutputCost:            outputCost,
 		CacheCreationCost:     cacheCreationCost,

backend/internal/repository/usage_log_repo_request_type_test.go

@@ -56,6 +56,8 @@ func TestUsageLogRepositoryCreateSyncRequestTypeAndLegacyFields(t *testing.T) {
 			log.CacheReadTokens,
 			log.CacheCreation5mTokens,
 			log.CacheCreation1hTokens,
+			log.ImageOutputTokens,
+			log.ImageOutputCost,
 			log.InputCost,
 			log.OutputCost,
 			log.CacheCreationCost,
@@ -133,6 +135,8 @@ func TestUsageLogRepositoryCreate_PersistsServiceTier(t *testing.T) {
 			log.CacheReadTokens,
 			log.CacheCreation5mTokens,
 			log.CacheCreation1hTokens,
+			log.ImageOutputTokens,
+			log.ImageOutputCost,
 			log.InputCost,
 			log.OutputCost,
 			log.CacheCreationCost,
@@ -447,6 +451,8 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			4,                 // cache_read_tokens
 			5,                 // cache_creation_5m_tokens
 			6,                 // cache_creation_1h_tokens
+			0,                 // image_output_tokens
+			0.0,               // image_output_cost
 			0.1,               // input_cost
 			0.2,               // output_cost
 			0.3,               // cache_creation_cost
@@ -499,6 +505,7 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			sql.NullInt64{},
 			sql.NullInt64{},
 			1, 2, 3, 4, 5, 6,
+			0, 0.0, // image_output_tokens, image_output_cost
 			0.1, 0.2, 0.3, 0.4, 1.0, 0.9,
 			1.0,
 			sql.NullFloat64{},
@@ -546,6 +553,7 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			sql.NullInt64{},
 			sql.NullInt64{},
 			1, 2, 3, 4, 5, 6,
+			0, 0.0, // image_output_tokens, image_output_cost
 			0.1, 0.2, 0.3, 0.4, 1.0, 0.9,
 			1.0,
 			sql.NullFloat64{},

backend/internal/service/account_usage_service.go

@@ -846,6 +846,15 @@ 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)
+}
+
 // 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) {

backend/internal/service/antigravity_credits_overages.go

@@ -19,6 +19,54 @@ const (
 	creditsExhaustedDuration = 5 * time.Hour
 )
 
+// 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。
+func (s *AntigravityGatewayService) checkAccountCredits(
+	ctx context.Context, account *Account, accessToken, proxyURL string,
+) 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
+	}
+
+	if usageInfo == nil || len(usageInfo.AICredits) == 0 {
+		logger.LegacyPrintf("service.antigravity_gateway",
+			"check_credits: account=%d has_credits=false amount=0 (no credits field)",
+			account.ID)
+		return false
+	}
+
+	for _, credit := range usageInfo.AICredits {
+		if credit.CreditType == "GOOGLE_ONE_AI" {
+			minimum := credit.MinimumBalance
+			if minimum <= 0 {
+				minimum = 5
+			}
+			hasCredits := credit.Amount >= minimum
+			logger.LegacyPrintf("service.antigravity_gateway",
+				"check_credits: account=%d has_credits=%t amount=%.0f minimum=%.0f",
+				account.ID, hasCredits, credit.Amount, minimum)
+			return hasCredits
+		}
+	}
+
+	logger.LegacyPrintf("service.antigravity_gateway",
+		"check_credits: account=%d has_credits=false (no GOOGLE_ONE_AI credit)",
+		account.ID)
+	return false
+}
+
 type antigravity429Category string
 
 const (
@@ -141,6 +189,13 @@ 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 + 非单模型限流：积分注入后仍 429 → 标记耗尽。
+//   - 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。
+//   - 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
@@ -148,13 +203,16 @@ func shouldMarkCreditsExhausted(resp *http.Response, respBody []byte, reqErr err
 	if resp.StatusCode >= 500 || resp.StatusCode == http.StatusRequestTimeout {
 		return false
 	}
-	// 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 {
+	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
 		}
-	bodyLower := strings.ToLower(string(respBody))
+		return true
+	}
+	// 其他 4xx：关键词匹配（如 403 + "Insufficient credits"）
 	for _, keyword := range creditsExhaustedKeywords {
 		if strings.Contains(bodyLower, keyword) {
 			return true
@@ -181,6 +239,16 @@ 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, p.accessToken, p.proxyURL) {
+		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)

backend/internal/service/antigravity_credits_overages_test.go

@@ -418,7 +418,13 @@ func TestShouldMarkCreditsExhausted(t *testing.T) {
 		require.True(t, shouldMarkCreditsExhausted(resp, body, nil))
 	})
 
-	t.Run("结构化限流不标记", func(t *testing.T) {
+	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 结构化限流也标记（积分注入后仍 429 即为耗尽）", 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))

backend/internal/service/antigravity_gateway_service.go

@@ -557,7 +557,13 @@ 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) {
-		if creditsBody := injectEnabledCreditTypes(p.body); creditsBody != nil {
+		// Check actual credits balance before injection
+		if !s.checkAccountCredits(p.ctx, p.account, p.accessToken, p.proxyURL) {
+			// 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 {
 			p.body = creditsBody
 			overagesInjected = true
 			logger.LegacyPrintf("service.antigravity_gateway", "%s pre_check: model_rate_limited_credits_inject model=%s account=%d (injecting enabledCreditTypes)",
@@ -878,6 +884,7 @@ type AntigravityGatewayService struct {
 	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
 }
 
 func NewAntigravityGatewayService(
@@ -889,6 +896,7 @@ func NewAntigravityGatewayService(
 	httpUpstream HTTPUpstream,
 	settingService *SettingService,
 	internal500Cache Internal500CounterCache,
+	accountUsageService *AccountUsageService,
 ) *AntigravityGatewayService {
 	return &AntigravityGatewayService{
 		accountRepo:         accountRepo,
@@ -899,6 +907,7 @@ func NewAntigravityGatewayService(
 		cache:               cache,
 		schedulerSnapshot:   schedulerSnapshot,
 		internal500Cache:    internal500Cache,
+		accountUsageService: accountUsageService,
 	}
 }
 

backend/internal/service/billing_service.go

@@ -56,6 +56,7 @@ type ModelPricing struct {
 	LongContextInputThreshold      int     // 超过阈值后按整次会话提升输入价格
 	LongContextInputMultiplier     float64 // 长上下文整次会话输入倍率
 	LongContextOutputMultiplier    float64 // 长上下文整次会话输出倍率
+	ImageOutputPricePerToken       float64 // 图片输出 token 价格 (USD)
 }
 
 const (
@@ -94,12 +95,14 @@ type UsageTokens struct {
 	CacheReadTokens       int
 	CacheCreation5mTokens int
 	CacheCreation1hTokens int
+	ImageOutputTokens     int
 }
 
 // CostBreakdown 费用明细
 type CostBreakdown struct {
 	InputCost         float64
 	OutputCost        float64
+	ImageOutputCost   float64
 	CacheCreationCost float64
 	CacheReadCost     float64
 	TotalCost         float64
@@ -358,6 +361,7 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 				LongContextInputThreshold:      litellmPricing.LongContextInputTokenThreshold,
 				LongContextInputMultiplier:     litellmPricing.LongContextInputCostMultiplier,
 				LongContextOutputMultiplier:    litellmPricing.LongContextOutputCostMultiplier,
+				ImageOutputPricePerToken:       litellmPricing.OutputCostPerImageToken,
 			}), nil
 		}
 	}
@@ -399,6 +403,9 @@ func (s *BillingService) GetModelPricingWithChannel(model string, channelPricing
 		pricing.CacheReadPricePerToken = *channelPricing.CacheReadPrice
 		pricing.CacheReadPricePerTokenPriority = *channelPricing.CacheReadPrice
 	}
+	if channelPricing.ImageOutputPrice != nil {
+		pricing.ImageOutputPricePerToken = *channelPricing.ImageOutputPrice
+	}
 	return pricing, nil
 }
 
@@ -489,7 +496,22 @@ func (s *BillingService) calculateTokenCost(resolved *ResolvedPricing, input Cos
 	}
 
 	breakdown.InputCost = float64(input.Tokens.InputTokens) * inputPricePerToken
-	breakdown.OutputCost = float64(input.Tokens.OutputTokens) * outputPricePerToken
+
+	// Separate image output tokens from text output tokens
+	textOutputTokens := input.Tokens.OutputTokens - input.Tokens.ImageOutputTokens
+	if textOutputTokens < 0 {
+		textOutputTokens = 0
+	}
+	breakdown.OutputCost = float64(textOutputTokens) * outputPricePerToken
+
+	// Image output tokens cost (separate rate from text output)
+	if input.Tokens.ImageOutputTokens > 0 {
+		imageOutputPrice := pricing.ImageOutputPricePerToken
+		if imageOutputPrice == 0 {
+			imageOutputPrice = outputPricePerToken // fallback to regular output price
+		}
+		breakdown.ImageOutputCost = float64(input.Tokens.ImageOutputTokens) * imageOutputPrice
+	}
 
 	if pricing.SupportsCacheBreakdown && (pricing.CacheCreation5mPrice > 0 || pricing.CacheCreation1hPrice > 0) {
 		if input.Tokens.CacheCreation5mTokens == 0 && input.Tokens.CacheCreation1hTokens == 0 && input.Tokens.CacheCreationTokens > 0 {
@@ -507,11 +529,12 @@ func (s *BillingService) calculateTokenCost(resolved *ResolvedPricing, input Cos
 	if tierMultiplier != 1.0 {
 		breakdown.InputCost *= tierMultiplier
 		breakdown.OutputCost *= tierMultiplier
+		breakdown.ImageOutputCost *= tierMultiplier
 		breakdown.CacheCreationCost *= tierMultiplier
 		breakdown.CacheReadCost *= tierMultiplier
 	}
 
-	breakdown.TotalCost = breakdown.InputCost + breakdown.OutputCost +
+	breakdown.TotalCost = breakdown.InputCost + breakdown.OutputCost + breakdown.ImageOutputCost +
 		breakdown.CacheCreationCost + breakdown.CacheReadCost
 	breakdown.ActualCost = breakdown.TotalCost * input.RateMultiplier
 
@@ -597,8 +620,21 @@ func (s *BillingService) calculateCostInternal(model string, tokens UsageTokens,
 	// 计算输入token费用（使用per-token价格）
 	breakdown.InputCost = float64(tokens.InputTokens) * inputPricePerToken
 
-	// 计算输出token费用
-	breakdown.OutputCost = float64(tokens.OutputTokens) * outputPricePerToken
+	// 计算输出token费用（分离图片输出token）
+	textOutputTokens := tokens.OutputTokens - tokens.ImageOutputTokens
+	if textOutputTokens < 0 {
+		textOutputTokens = 0
+	}
+	breakdown.OutputCost = float64(textOutputTokens) * outputPricePerToken
+
+	// 图片输出 token 费用
+	if tokens.ImageOutputTokens > 0 {
+		imageOutputPrice := pricing.ImageOutputPricePerToken
+		if imageOutputPrice == 0 {
+			imageOutputPrice = outputPricePerToken
+		}
+		breakdown.ImageOutputCost = float64(tokens.ImageOutputTokens) * imageOutputPrice
+	}
 
 	// 计算缓存费用
 	if pricing.SupportsCacheBreakdown && (pricing.CacheCreation5mPrice > 0 || pricing.CacheCreation1hPrice > 0) {
@@ -620,12 +656,13 @@ func (s *BillingService) calculateCostInternal(model string, tokens UsageTokens,
 	if tierMultiplier != 1.0 {
 		breakdown.InputCost *= tierMultiplier
 		breakdown.OutputCost *= tierMultiplier
+		breakdown.ImageOutputCost *= tierMultiplier
 		breakdown.CacheCreationCost *= tierMultiplier
 		breakdown.CacheReadCost *= tierMultiplier
 	}
 
 	// 计算总费用
-	breakdown.TotalCost = breakdown.InputCost + breakdown.OutputCost +
+	breakdown.TotalCost = breakdown.InputCost + breakdown.OutputCost + breakdown.ImageOutputCost +
 		breakdown.CacheCreationCost + breakdown.CacheReadCost
 
 	// 应用倍率计算实际费用
@@ -730,6 +767,7 @@ func (s *BillingService) CalculateCostWithLongContext(model string, tokens Usage
 		CacheReadTokens:       inRangeCacheTokens,
 		CacheCreation5mTokens: tokens.CacheCreation5mTokens,
 		CacheCreation1hTokens: tokens.CacheCreation1hTokens,
+		ImageOutputTokens:     tokens.ImageOutputTokens,
 	}
 	inRangeCost, err := s.CalculateCost(model, inRangeTokens, rateMultiplier)
 	if err != nil {
@@ -750,6 +788,7 @@ func (s *BillingService) CalculateCostWithLongContext(model string, tokens Usage
 	return &CostBreakdown{
 		InputCost:         inRangeCost.InputCost + outRangeCost.InputCost,
 		OutputCost:        inRangeCost.OutputCost,
+		ImageOutputCost:   inRangeCost.ImageOutputCost,
 		CacheCreationCost: inRangeCost.CacheCreationCost,
 		CacheReadCost:     inRangeCost.CacheReadCost + outRangeCost.CacheReadCost,
 		TotalCost:         inRangeCost.TotalCost + outRangeCost.TotalCost,

backend/internal/service/channel.go

@@ -26,6 +26,7 @@ func (m BillingMode) IsValid() bool {
 const (
 	BillingModelSourceRequested     = "requested"
 	BillingModelSourceUpstream      = "upstream"
+	BillingModelSourceChannelMapped = "channel_mapped"
 )
 
 // Channel 渠道实体
@@ -34,7 +35,7 @@ type Channel struct {
 	Name               string
 	Description        string
 	Status             string
-	BillingModelSource string // "requested" or "upstream"
+	BillingModelSource string // "requested", "upstream", or "channel_mapped"
 	RestrictModels     bool   // 是否限制模型（仅允许定价列表中的模型）
 	CreatedAt          time.Time
 	UpdatedAt          time.Time
@@ -180,6 +181,7 @@ func (c *Channel) Clone() *Channel {
 type ChannelUsageFields struct {
 	ChannelID          int64  // 渠道 ID（0 = 无渠道）
 	OriginalModel      string // 用户原始请求模型（渠道映射前）
-	BillingModelSource string // 计费模型来源："requested" / "upstream"
+	ChannelMappedModel string // 渠道映射后的模型名（无映射时等于 OriginalModel）
+	BillingModelSource string // 计费模型来源："requested" / "upstream" / "channel_mapped"
 	ModelMappingChain  string // 映射链描述，如 "a→b→c"
 }

backend/internal/service/channel_service.go

@@ -97,7 +97,7 @@ type ChannelMappingResult struct {
 	MappedModel        string // 映射后的模型名（无映射时等于原始模型名）
 	ChannelID          int64  // 渠道 ID（0 = 无渠道关联）
 	Mapped             bool   // 是否发生了映射
-	BillingModelSource string // 计费模型来源（"requested" / "upstream"）
+	BillingModelSource string // 计费模型来源（"requested" / "upstream" / "channel_mapped"）
 }
 
 // BuildModelMappingChain 根据映射结果和上游实际模型构建映射链描述。
@@ -119,9 +119,14 @@ func (r ChannelMappingResult) BuildModelMappingChain(reqModel, upstreamModel str
 
 // ToUsageFields 将渠道映射结果转为使用记录字段
 func (r ChannelMappingResult) ToUsageFields(reqModel, upstreamModel string) ChannelUsageFields {
+	channelMappedModel := reqModel
+	if r.Mapped {
+		channelMappedModel = r.MappedModel
+	}
 	return ChannelUsageFields{
 		ChannelID:          r.ChannelID,
 		OriginalModel:      reqModel,
+		ChannelMappedModel: channelMappedModel,
 		BillingModelSource: r.BillingModelSource,
 		ModelMappingChain:  r.BuildModelMappingChain(reqModel, upstreamModel),
 	}
@@ -193,7 +198,7 @@ func (s *ChannelService) buildCache(ctx context.Context) (*channelCache, error)
 			channelByGroupID:        make(map[int64]*Channel),
 			groupPlatform:           make(map[int64]string),
 			byID:                    make(map[int64]*Channel),
-			loadedAt:                time.Now().Add(channelCacheTTL - channelErrorTTL), // 使剩余 TTL = errorTTL
+			loadedAt:                time.Now().Add(-(channelCacheTTL - channelErrorTTL)), // 使剩余 TTL = errorTTL
 		}
 		s.cache.Store(errorCache)
 		return nil, fmt.Errorf("list all channels: %w", err)
@@ -374,7 +379,7 @@ func (s *ChannelService) ResolveChannelMapping(ctx context.Context, groupID int6
 		BillingModelSource: ch.BillingModelSource,
 	}
 	if result.BillingModelSource == "" {
-		result.BillingModelSource = BillingModelSourceRequested
+		result.BillingModelSource = BillingModelSourceChannelMapped
 	}
 
 	platform := cache.groupPlatform[groupID]
@@ -481,7 +486,7 @@ func (s *ChannelService) Create(ctx context.Context, input *CreateChannelInput)
 		ModelMapping:       input.ModelMapping,
 	}
 	if channel.BillingModelSource == "" {
-		channel.BillingModelSource = BillingModelSourceRequested
+		channel.BillingModelSource = BillingModelSourceChannelMapped
 	}
 
 	if err := validateNoConflictingModels(channel.ModelPricing); err != nil {
@@ -565,22 +570,38 @@ func (s *ChannelService) Update(ctx context.Context, id int64, input *UpdateChan
 		return nil, err
 	}
 
+	// 先获取旧分组，Update 后旧分组关联已删除，无法再查到
+	var oldGroupIDs []int64
+	if s.authCacheInvalidator != nil {
+		var err2 error
+		oldGroupIDs, err2 = s.repo.GetGroupIDs(ctx, id)
+		if err2 != nil {
+			slog.Warn("failed to get old group IDs for cache invalidation", "channel_id", id, "error", err2)
+		}
+	}
+
 	if err := s.repo.Update(ctx, channel); err != nil {
 		return nil, fmt.Errorf("update channel: %w", err)
 	}
 
 	s.invalidateCache()
 
-	// 失效关联分组的 auth 缓存
+	// 失效新旧分组的 auth 缓存
 	if s.authCacheInvalidator != nil {
-		groupIDs, err := s.repo.GetGroupIDs(ctx, id)
-		if err != nil {
-			slog.Warn("failed to get group IDs for cache invalidation", "channel_id", id, "error", err)
+		seen := make(map[int64]struct{}, len(oldGroupIDs)+len(channel.GroupIDs))
+		for _, gid := range oldGroupIDs {
+			if _, ok := seen[gid]; !ok {
+				seen[gid] = struct{}{}
+				s.authCacheInvalidator.InvalidateAuthCacheByGroupID(ctx, gid)
 			}
-		for _, gid := range groupIDs {
+		}
+		for _, gid := range channel.GroupIDs {
+			if _, ok := seen[gid]; !ok {
+				seen[gid] = struct{}{}
 				s.authCacheInvalidator.InvalidateAuthCacheByGroupID(ctx, gid)
 			}
 		}
+	}
 
 	return s.repo.GetByID(ctx, id)
 }

backend/internal/service/channel_service_test.go

@@ -196,7 +196,6 @@ func newTestChannelServiceWithAuth(repo *mockChannelRepository, auth *mockChanne
 	return NewChannelService(repo, auth)
 }
 
-
 // makeStandardRepo returns a repo that serves one active channel with anthropic pricing
 // for group 1, with the given model pricing and model mapping.
 func makeStandardRepo(ch Channel, groupPlatforms map[int64]string) *mockChannelRepository {
@@ -914,7 +913,7 @@ func TestResolveChannelMapping_DefaultBillingModelSource(t *testing.T) {
 	svc := newTestChannelService(repo)
 
 	result := svc.ResolveChannelMapping(context.Background(), 10, "claude-opus-4")
-	require.Equal(t, BillingModelSourceRequested, result.BillingModelSource)
+	require.Equal(t, BillingModelSourceChannelMapped, result.BillingModelSource)
 }
 
 func TestResolveChannelMapping_UpstreamBillingModelSource(t *testing.T) {
@@ -1451,11 +1450,11 @@ func TestCreate_DefaultBillingModelSource(t *testing.T) {
 
 	result, err := svc.Create(context.Background(), &CreateChannelInput{
 		Name:               "new-channel",
-		BillingModelSource: "", // empty, should default to "requested"
+		BillingModelSource: "", // empty, should default to "channel_mapped"
 	})
 	require.NoError(t, err)
 	require.NotNil(t, result)
-	require.Equal(t, BillingModelSourceRequested, result.BillingModelSource)
+	require.Equal(t, BillingModelSourceChannelMapped, result.BillingModelSource)
 }
 
 func TestCreate_InvalidatesCache(t *testing.T) {

backend/internal/service/gateway_service.go

@@ -483,6 +483,7 @@ type ClaudeUsage struct {
 	CacheReadInputTokens     int `json:"cache_read_input_tokens"`
 	CacheCreation5mTokens    int // 5分钟缓存创建token（来自嵌套 cache_creation 对象）
 	CacheCreation1hTokens    int // 1小时缓存创建token（来自嵌套 cache_creation 对象）
+	ImageOutputTokens        int `json:"image_output_tokens,omitempty"`
 }
 
 // ForwardResult 转发结果
@@ -7729,6 +7730,9 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 	var cost *CostBreakdown
 	// 确定计费模型
 	billingModel := forwardResultBillingModel(result.Model, result.UpstreamModel)
+	if input.BillingModelSource == BillingModelSourceChannelMapped && input.ChannelMappedModel != "" {
+		billingModel = input.ChannelMappedModel
+	}
 	if input.BillingModelSource == BillingModelSourceRequested && input.OriginalModel != "" {
 		billingModel = input.OriginalModel
 	}
@@ -7777,6 +7781,7 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 			CacheReadTokens:       result.Usage.CacheReadInputTokens,
 			CacheCreation5mTokens: result.Usage.CacheCreation5mTokens,
 			CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
+			ImageOutputTokens:     result.Usage.ImageOutputTokens,
 		}
 		var err error
 		if s.resolver != nil && apiKey.Group != nil {
@@ -7836,8 +7841,10 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 		CacheReadTokens:       result.Usage.CacheReadInputTokens,
 		CacheCreation5mTokens: result.Usage.CacheCreation5mTokens,
 		CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
+		ImageOutputTokens:     result.Usage.ImageOutputTokens,
 		InputCost:             cost.InputCost,
 		OutputCost:            cost.OutputCost,
+		ImageOutputCost:       cost.ImageOutputCost,
 		CacheCreationCost:     cost.CacheCreationCost,
 		CacheReadCost:         cost.CacheReadCost,
 		TotalCost:             cost.TotalCost,
@@ -7976,6 +7983,9 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 	var cost *CostBreakdown
 	// 确定计费模型
 	billingModel := forwardResultBillingModel(result.Model, result.UpstreamModel)
+	if input.BillingModelSource == BillingModelSourceChannelMapped && input.ChannelMappedModel != "" {
+		billingModel = input.ChannelMappedModel
+	}
 	if input.BillingModelSource == BillingModelSourceRequested && input.OriginalModel != "" {
 		billingModel = input.OriginalModel
 	}
@@ -8007,6 +8017,7 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 			CacheReadTokens:       result.Usage.CacheReadInputTokens,
 			CacheCreation5mTokens: result.Usage.CacheCreation5mTokens,
 			CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
+			ImageOutputTokens:     result.Usage.ImageOutputTokens,
 		}
 		var err error
 		// 优先尝试 Resolver + CalculateCostUnified（仅在有渠道定价时使用）
@@ -8073,8 +8084,10 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 		CacheReadTokens:       result.Usage.CacheReadInputTokens,
 		CacheCreation5mTokens: result.Usage.CacheCreation5mTokens,
 		CacheCreation1hTokens: result.Usage.CacheCreation1hTokens,
+		ImageOutputTokens:     result.Usage.ImageOutputTokens,
 		InputCost:             cost.InputCost,
 		OutputCost:            cost.OutputCost,
+		ImageOutputCost:       cost.ImageOutputCost,
 		CacheCreationCost:     cost.CacheCreationCost,
 		CacheReadCost:         cost.CacheReadCost,
 		TotalCost:             cost.TotalCost,

backend/internal/service/gemini_messages_compat_service.go

@@ -2692,12 +2692,27 @@ func extractGeminiUsage(data []byte) *ClaudeUsage {
 	cand := int(usage.Get("candidatesTokenCount").Int())
 	cached := int(usage.Get("cachedContentTokenCount").Int())
 	thoughts := int(usage.Get("thoughtsTokenCount").Int())
+
+	// 从 candidatesTokensDetails 提取 IMAGE 模态 token 数
+	imageTokens := 0
+	candidateDetails := usage.Get("candidatesTokensDetails")
+	if candidateDetails.Exists() {
+		candidateDetails.ForEach(func(_, detail gjson.Result) bool {
+			if detail.Get("modality").String() == "IMAGE" {
+				imageTokens = int(detail.Get("tokenCount").Int())
+				return false
+			}
+			return true
+		})
+	}
+
 	// 注意：Gemini 的 promptTokenCount 包含 cachedContentTokenCount，
 	// 但 Claude 的 input_tokens 不包含 cache_read_input_tokens，需要减去
 	return &ClaudeUsage{
 		InputTokens:          prompt - cached,
 		OutputTokens:         cand + thoughts,
 		CacheReadInputTokens: cached,
+		ImageOutputTokens:    imageTokens,
 	}
 }