Merge pull request #800 from mt21625457/pr/gpt54-support-upstream

feat(openai): 增加 GPT-5.4 支持并修复长上下文计费与白名单回归

backend/internal/pkg/openai/constants.go

@@ -15,6 +15,7 @@ type Model struct {
 
 // DefaultModels OpenAI models list
 var DefaultModels = []Model{
+	{ID: "gpt-5.4", Object: "model", Created: 1738368000, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.4"},
 	{ID: "gpt-5.3-codex", Object: "model", Created: 1735689600, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.3 Codex"},
 	{ID: "gpt-5.3-codex-spark", Object: "model", Created: 1735689600, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.3 Codex Spark"},
 	{ID: "gpt-5.2", Object: "model", Created: 1733875200, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.2"},

backend/internal/service/billing_service.go

@@ -43,15 +43,24 @@ type BillingCache interface {
 
 // ModelPricing 模型价格配置（per-token价格，与LiteLLM格式一致）
 type ModelPricing struct {
-	InputPricePerToken         float64 // 每token输入价格 (USD)
-	OutputPricePerToken        float64 // 每token输出价格 (USD)
-	CacheCreationPricePerToken float64 // 缓存创建每token价格 (USD)
-	CacheReadPricePerToken     float64 // 缓存读取每token价格 (USD)
-	CacheCreation5mPrice       float64 // 5分钟缓存创建每token价格 (USD)
-	CacheCreation1hPrice       float64 // 1小时缓存创建每token价格 (USD)
-	SupportsCacheBreakdown     bool    // 是否支持详细的缓存分类
+	InputPricePerToken          float64 // 每token输入价格 (USD)
+	OutputPricePerToken         float64 // 每token输出价格 (USD)
+	CacheCreationPricePerToken  float64 // 缓存创建每token价格 (USD)
+	CacheReadPricePerToken      float64 // 缓存读取每token价格 (USD)
+	CacheCreation5mPrice        float64 // 5分钟缓存创建每token价格 (USD)
+	CacheCreation1hPrice        float64 // 1小时缓存创建每token价格 (USD)
+	SupportsCacheBreakdown      bool    // 是否支持详细的缓存分类
+	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
 }
 
+const (
+	openAIGPT54LongContextInputThreshold   = 272000
+	openAIGPT54LongContextInputMultiplier  = 2.0
+	openAIGPT54LongContextOutputMultiplier = 1.5
+)
+
 // UsageTokens 使用的token数量
 type UsageTokens struct {
 	InputTokens           int
@@ -161,6 +170,35 @@ func (s *BillingService) initFallbackPricing() {
 		CacheReadPricePerToken:     0.2e-6, // $0.20 per MTok
 		SupportsCacheBreakdown:     false,
 	}
+
+	// OpenAI GPT-5.1（本地兜底，防止动态定价不可用时拒绝计费）
+	s.fallbackPrices["gpt-5.1"] = &ModelPricing{
+		InputPricePerToken:         1.25e-6, // $1.25 per MTok
+		OutputPricePerToken:        10e-6,   // $10 per MTok
+		CacheCreationPricePerToken: 1.25e-6, // $1.25 per MTok
+		CacheReadPricePerToken:     0.125e-6,
+		SupportsCacheBreakdown:     false,
+	}
+	// OpenAI GPT-5.4（业务指定价格）
+	s.fallbackPrices["gpt-5.4"] = &ModelPricing{
+		InputPricePerToken:          2.5e-6,  // $2.5 per MTok
+		OutputPricePerToken:         15e-6,   // $15 per MTok
+		CacheCreationPricePerToken:  2.5e-6,  // $2.5 per MTok
+		CacheReadPricePerToken:      0.25e-6, // $0.25 per MTok
+		SupportsCacheBreakdown:      false,
+		LongContextInputThreshold:   openAIGPT54LongContextInputThreshold,
+		LongContextInputMultiplier:  openAIGPT54LongContextInputMultiplier,
+		LongContextOutputMultiplier: openAIGPT54LongContextOutputMultiplier,
+	}
+	// Codex 族兜底统一按 GPT-5.1 Codex 价格计费
+	s.fallbackPrices["gpt-5.1-codex"] = &ModelPricing{
+		InputPricePerToken:         1.5e-6, // $1.5 per MTok
+		OutputPricePerToken:        12e-6,  // $12 per MTok
+		CacheCreationPricePerToken: 1.5e-6, // $1.5 per MTok
+		CacheReadPricePerToken:     0.15e-6,
+		SupportsCacheBreakdown:     false,
+	}
+	s.fallbackPrices["gpt-5.3-codex"] = s.fallbackPrices["gpt-5.1-codex"]
 }
 
 // getFallbackPricing 根据模型系列获取回退价格
@@ -189,12 +227,30 @@ func (s *BillingService) getFallbackPricing(model string) *ModelPricing {
 		}
 		return s.fallbackPrices["claude-3-haiku"]
 	}
+	// Claude 未知型号统一回退到 Sonnet，避免计费中断。
+	if strings.Contains(modelLower, "claude") {
+		return s.fallbackPrices["claude-sonnet-4"]
+	}
 	if strings.Contains(modelLower, "gemini-3.1-pro") || strings.Contains(modelLower, "gemini-3-1-pro") {
 		return s.fallbackPrices["gemini-3.1-pro"]
 	}
 
-	// 默认使用Sonnet价格
-	return s.fallbackPrices["claude-sonnet-4"]
+	// OpenAI 仅匹配已知 GPT-5/Codex 族，避免未知 OpenAI 型号误计价。
+	if strings.Contains(modelLower, "gpt-5") || strings.Contains(modelLower, "codex") {
+		normalized := normalizeCodexModel(modelLower)
+		switch normalized {
+		case "gpt-5.4":
+			return s.fallbackPrices["gpt-5.4"]
+		case "gpt-5.3-codex":
+			return s.fallbackPrices["gpt-5.3-codex"]
+		case "gpt-5.1-codex", "gpt-5.1-codex-max", "gpt-5.1-codex-mini", "codex-mini-latest":
+			return s.fallbackPrices["gpt-5.1-codex"]
+		case "gpt-5.1":
+			return s.fallbackPrices["gpt-5.1"]
+		}
+	}
+
+	return nil
 }
 
 // GetModelPricing 获取模型价格配置
@@ -212,15 +268,18 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 			price5m := litellmPricing.CacheCreationInputTokenCost
 			price1h := litellmPricing.CacheCreationInputTokenCostAbove1hr
 			enableBreakdown := price1h > 0 && price1h > price5m
-			return &ModelPricing{
-				InputPricePerToken:         litellmPricing.InputCostPerToken,
-				OutputPricePerToken:        litellmPricing.OutputCostPerToken,
-				CacheCreationPricePerToken: litellmPricing.CacheCreationInputTokenCost,
-				CacheReadPricePerToken:     litellmPricing.CacheReadInputTokenCost,
-				CacheCreation5mPrice:       price5m,
-				CacheCreation1hPrice:       price1h,
-				SupportsCacheBreakdown:     enableBreakdown,
-			}, nil
+			return s.applyModelSpecificPricingPolicy(model, &ModelPricing{
+				InputPricePerToken:          litellmPricing.InputCostPerToken,
+				OutputPricePerToken:         litellmPricing.OutputCostPerToken,
+				CacheCreationPricePerToken:  litellmPricing.CacheCreationInputTokenCost,
+				CacheReadPricePerToken:      litellmPricing.CacheReadInputTokenCost,
+				CacheCreation5mPrice:        price5m,
+				CacheCreation1hPrice:        price1h,
+				SupportsCacheBreakdown:      enableBreakdown,
+				LongContextInputThreshold:   litellmPricing.LongContextInputTokenThreshold,
+				LongContextInputMultiplier:  litellmPricing.LongContextInputCostMultiplier,
+				LongContextOutputMultiplier: litellmPricing.LongContextOutputCostMultiplier,
+			}), nil
 		}
 	}
 
@@ -228,7 +287,7 @@ func (s *BillingService) GetModelPricing(model string) (*ModelPricing, error) {
 	fallback := s.getFallbackPricing(model)
 	if fallback != nil {
 		log.Printf("[Billing] Using fallback pricing for model: %s", model)
-		return fallback, nil
+		return s.applyModelSpecificPricingPolicy(model, fallback), nil
 	}
 
 	return nil, fmt.Errorf("pricing not found for model: %s", model)
@@ -242,12 +301,18 @@ func (s *BillingService) CalculateCost(model string, tokens UsageTokens, rateMul
 	}
 
 	breakdown := &CostBreakdown{}
+	inputPricePerToken := pricing.InputPricePerToken
+	outputPricePerToken := pricing.OutputPricePerToken
+	if s.shouldApplySessionLongContextPricing(tokens, pricing) {
+		inputPricePerToken *= pricing.LongContextInputMultiplier
+		outputPricePerToken *= pricing.LongContextOutputMultiplier
+	}
 
 	// 计算输入token费用（使用per-token价格）
-	breakdown.InputCost = float64(tokens.InputTokens) * pricing.InputPricePerToken
+	breakdown.InputCost = float64(tokens.InputTokens) * inputPricePerToken
 
 	// 计算输出token费用
-	breakdown.OutputCost = float64(tokens.OutputTokens) * pricing.OutputPricePerToken
+	breakdown.OutputCost = float64(tokens.OutputTokens) * outputPricePerToken
 
 	// 计算缓存费用
 	if pricing.SupportsCacheBreakdown && (pricing.CacheCreation5mPrice > 0 || pricing.CacheCreation1hPrice > 0) {
@@ -279,6 +344,45 @@ func (s *BillingService) CalculateCost(model string, tokens UsageTokens, rateMul
 	return breakdown, nil
 }
 
+func (s *BillingService) applyModelSpecificPricingPolicy(model string, pricing *ModelPricing) *ModelPricing {
+	if pricing == nil {
+		return nil
+	}
+	if !isOpenAIGPT54Model(model) {
+		return pricing
+	}
+	if pricing.LongContextInputThreshold > 0 && pricing.LongContextInputMultiplier > 0 && pricing.LongContextOutputMultiplier > 0 {
+		return pricing
+	}
+	cloned := *pricing
+	if cloned.LongContextInputThreshold <= 0 {
+		cloned.LongContextInputThreshold = openAIGPT54LongContextInputThreshold
+	}
+	if cloned.LongContextInputMultiplier <= 0 {
+		cloned.LongContextInputMultiplier = openAIGPT54LongContextInputMultiplier
+	}
+	if cloned.LongContextOutputMultiplier <= 0 {
+		cloned.LongContextOutputMultiplier = openAIGPT54LongContextOutputMultiplier
+	}
+	return &cloned
+}
+
+func (s *BillingService) shouldApplySessionLongContextPricing(tokens UsageTokens, pricing *ModelPricing) bool {
+	if pricing == nil || pricing.LongContextInputThreshold <= 0 {
+		return false
+	}
+	if pricing.LongContextInputMultiplier <= 1 && pricing.LongContextOutputMultiplier <= 1 {
+		return false
+	}
+	totalInputTokens := tokens.InputTokens + tokens.CacheReadTokens
+	return totalInputTokens > pricing.LongContextInputThreshold
+}
+
+func isOpenAIGPT54Model(model string) bool {
+	normalized := normalizeCodexModel(strings.TrimSpace(strings.ToLower(model)))
+	return normalized == "gpt-5.4"
+}
+
 // CalculateCostWithConfig 使用配置中的默认倍率计算费用
 func (s *BillingService) CalculateCostWithConfig(model string, tokens UsageTokens) (*CostBreakdown, error) {
 	multiplier := s.cfg.Default.RateMultiplier

backend/internal/service/billing_service_test.go

@@ -133,7 +133,7 @@ func TestGetModelPricing_CaseInsensitive(t *testing.T) {
 	require.Equal(t, p1.InputPricePerToken, p2.InputPricePerToken)
 }
 
-func TestGetModelPricing_UnknownModelFallsBackToSonnet(t *testing.T) {
+func TestGetModelPricing_UnknownClaudeModelFallsBackToSonnet(t *testing.T) {
 	svc := newTestBillingService()
 
 	// 不包含 opus/sonnet/haiku 关键词的 Claude 模型会走默认 Sonnet 价格
@@ -142,6 +142,93 @@ func TestGetModelPricing_UnknownModelFallsBackToSonnet(t *testing.T) {
 	require.InDelta(t, 3e-6, pricing.InputPricePerToken, 1e-12)
 }
 
+func TestGetModelPricing_UnknownOpenAIModelReturnsError(t *testing.T) {
+	svc := newTestBillingService()
+
+	pricing, err := svc.GetModelPricing("gpt-unknown-model")
+	require.Error(t, err)
+	require.Nil(t, pricing)
+	require.Contains(t, err.Error(), "pricing not found")
+}
+
+func TestGetModelPricing_OpenAIGPT51Fallback(t *testing.T) {
+	svc := newTestBillingService()
+
+	pricing, err := svc.GetModelPricing("gpt-5.1")
+	require.NoError(t, err)
+	require.NotNil(t, pricing)
+	require.InDelta(t, 1.25e-6, pricing.InputPricePerToken, 1e-12)
+}
+
+func TestGetModelPricing_OpenAIGPT54Fallback(t *testing.T) {
+	svc := newTestBillingService()
+
+	pricing, err := svc.GetModelPricing("gpt-5.4")
+	require.NoError(t, err)
+	require.NotNil(t, pricing)
+	require.InDelta(t, 2.5e-6, pricing.InputPricePerToken, 1e-12)
+	require.InDelta(t, 15e-6, pricing.OutputPricePerToken, 1e-12)
+	require.InDelta(t, 0.25e-6, pricing.CacheReadPricePerToken, 1e-12)
+	require.Equal(t, 272000, pricing.LongContextInputThreshold)
+	require.InDelta(t, 2.0, pricing.LongContextInputMultiplier, 1e-12)
+	require.InDelta(t, 1.5, pricing.LongContextOutputMultiplier, 1e-12)
+}
+
+func TestCalculateCost_OpenAIGPT54LongContextAppliesWholeSessionMultipliers(t *testing.T) {
+	svc := newTestBillingService()
+
+	tokens := UsageTokens{
+		InputTokens:  300000,
+		OutputTokens: 4000,
+	}
+
+	cost, err := svc.CalculateCost("gpt-5.4-2026-03-05", tokens, 1.0)
+	require.NoError(t, err)
+
+	expectedInput := float64(tokens.InputTokens) * 2.5e-6 * 2.0
+	expectedOutput := float64(tokens.OutputTokens) * 15e-6 * 1.5
+	require.InDelta(t, expectedInput, cost.InputCost, 1e-10)
+	require.InDelta(t, expectedOutput, cost.OutputCost, 1e-10)
+	require.InDelta(t, expectedInput+expectedOutput, cost.TotalCost, 1e-10)
+	require.InDelta(t, expectedInput+expectedOutput, cost.ActualCost, 1e-10)
+}
+
+func TestGetFallbackPricing_FamilyMatching(t *testing.T) {
+	svc := newTestBillingService()
+
+	tests := []struct {
+		name             string
+		model            string
+		expectedInput    float64
+		expectNilPricing bool
+	}{
+		{name: "empty model", model: "   ", expectNilPricing: true},
+		{name: "claude opus 4.6", model: "claude-opus-4.6-20260201", expectedInput: 5e-6},
+		{name: "claude opus 4.5 alt separator", model: "claude-opus-4-5-20260101", expectedInput: 5e-6},
+		{name: "claude generic model fallback sonnet", model: "claude-foo-bar", expectedInput: 3e-6},
+		{name: "gemini explicit fallback", model: "gemini-3-1-pro", expectedInput: 2e-6},
+		{name: "gemini unknown no fallback", model: "gemini-2.0-pro", expectNilPricing: true},
+		{name: "openai gpt5.1", model: "gpt-5.1", expectedInput: 1.25e-6},
+		{name: "openai gpt5.4", model: "gpt-5.4", expectedInput: 2.5e-6},
+		{name: "openai gpt5.3 codex", model: "gpt-5.3-codex", expectedInput: 1.5e-6},
+		{name: "openai gpt5.1 codex max alias", model: "gpt-5.1-codex-max", expectedInput: 1.5e-6},
+		{name: "openai codex mini latest alias", model: "codex-mini-latest", expectedInput: 1.5e-6},
+		{name: "openai unknown no fallback", model: "gpt-unknown-model", expectNilPricing: true},
+		{name: "non supported family", model: "qwen-max", expectNilPricing: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			pricing := svc.getFallbackPricing(tt.model)
+			if tt.expectNilPricing {
+				require.Nil(t, pricing)
+				return
+			}
+			require.NotNil(t, pricing)
+			require.InDelta(t, tt.expectedInput, pricing.InputPricePerToken, 1e-12)
+		})
+	}
+}
 func TestCalculateCostWithLongContext_BelowThreshold(t *testing.T) {
 	svc := newTestBillingService()
 

backend/internal/service/openai_codex_transform.go

@@ -9,6 +9,13 @@ import (
 var codexCLIInstructions string
 
 var codexModelMap = map[string]string{
+	"gpt-5.4":                    "gpt-5.4",
+	"gpt-5.4-none":               "gpt-5.4",
+	"gpt-5.4-low":                "gpt-5.4",
+	"gpt-5.4-medium":             "gpt-5.4",
+	"gpt-5.4-high":               "gpt-5.4",
+	"gpt-5.4-xhigh":              "gpt-5.4",
+	"gpt-5.4-chat-latest":        "gpt-5.4",
 	"gpt-5.3":                    "gpt-5.3-codex",
 	"gpt-5.3-none":               "gpt-5.3-codex",
 	"gpt-5.3-low":                "gpt-5.3-codex",
@@ -154,6 +161,9 @@ func normalizeCodexModel(model string) string {
 
 	normalized := strings.ToLower(modelID)
 
+	if strings.Contains(normalized, "gpt-5.4") || strings.Contains(normalized, "gpt 5.4") {
+		return "gpt-5.4"
+	}
 	if strings.Contains(normalized, "gpt-5.2-codex") || strings.Contains(normalized, "gpt 5.2 codex") {
 		return "gpt-5.2-codex"
 	}

backend/internal/service/openai_codex_transform_test.go

@@ -167,6 +167,10 @@ func TestApplyCodexOAuthTransform_EmptyInput(t *testing.T) {
 
 func TestNormalizeCodexModel_Gpt53(t *testing.T) {
 	cases := map[string]string{
+		"gpt-5.4":                   "gpt-5.4",
+		"gpt-5.4-high":              "gpt-5.4",
+		"gpt-5.4-chat-latest":       "gpt-5.4",
+		"gpt 5.4":                   "gpt-5.4",
 		"gpt-5.3":                   "gpt-5.3-codex",
 		"gpt-5.3-codex":             "gpt-5.3-codex",
 		"gpt-5.3-codex-xhigh":       "gpt-5.3-codex",

backend/internal/service/pricing_service.go

@@ -21,8 +21,19 @@ import (
 )
 
 var (
-	openAIModelDatePattern = regexp.MustCompile(`-\d{8}$`)
-	openAIModelBasePattern = regexp.MustCompile(`^(gpt-\d+(?:\.\d+)?)(?:-|$)`)
+	openAIModelDatePattern     = regexp.MustCompile(`-\d{8}$`)
+	openAIModelBasePattern     = regexp.MustCompile(`^(gpt-\d+(?:\.\d+)?)(?:-|$)`)
+	openAIGPT54FallbackPricing = &LiteLLMModelPricing{
+		InputCostPerToken:               2.5e-06, // $2.5 per MTok
+		OutputCostPerToken:              1.5e-05, // $15 per MTok
+		CacheReadInputTokenCost:         2.5e-07, // $0.25 per MTok
+		LongContextInputTokenThreshold:  272000,
+		LongContextInputCostMultiplier:  2.0,
+		LongContextOutputCostMultiplier: 1.5,
+		LiteLLMProvider:                 "openai",
+		Mode:                            "chat",
+		SupportsPromptCaching:           true,
+	}
 )
 
 // LiteLLMModelPricing LiteLLM价格数据结构
@@ -33,6 +44,9 @@ type LiteLLMModelPricing struct {
 	CacheCreationInputTokenCost         float64 `json:"cache_creation_input_token_cost"`
 	CacheCreationInputTokenCostAbove1hr float64 `json:"cache_creation_input_token_cost_above_1hr"`
 	CacheReadInputTokenCost             float64 `json:"cache_read_input_token_cost"`
+	LongContextInputTokenThreshold      int     `json:"long_context_input_token_threshold,omitempty"`
+	LongContextInputCostMultiplier      float64 `json:"long_context_input_cost_multiplier,omitempty"`
+	LongContextOutputCostMultiplier     float64 `json:"long_context_output_cost_multiplier,omitempty"`
 	LiteLLMProvider                     string  `json:"litellm_provider"`
 	Mode                                string  `json:"mode"`
 	SupportsPromptCaching               bool    `json:"supports_prompt_caching"`
@@ -660,7 +674,8 @@ func (s *PricingService) matchByModelFamily(model string) *LiteLLMModelPricing {
 // 2. gpt-5.2-codex -> gpt-5.2（去掉后缀如 -codex, -mini, -max 等）
 // 3. gpt-5.2-20251222 -> gpt-5.2（去掉日期版本号）
 // 4. gpt-5.3-codex -> gpt-5.2-codex
-// 5. 最终回退到 DefaultTestModel (gpt-5.1-codex)
+// 5. gpt-5.4* -> 业务静态兜底价
+// 6. 最终回退到 DefaultTestModel (gpt-5.1-codex)
 func (s *PricingService) matchOpenAIModel(model string) *LiteLLMModelPricing {
 	if strings.HasPrefix(model, "gpt-5.3-codex-spark") {
 		if pricing, ok := s.pricingData["gpt-5.1-codex"]; ok {
@@ -690,6 +705,12 @@ func (s *PricingService) matchOpenAIModel(model string) *LiteLLMModelPricing {
 		}
 	}
 
+	if strings.HasPrefix(model, "gpt-5.4") {
+		logger.With(zap.String("component", "service.pricing")).
+			Info(fmt.Sprintf("[Pricing] OpenAI fallback matched %s -> %s", model, "gpt-5.4(static)"))
+		return openAIGPT54FallbackPricing
+	}
+
 	// 最终回退到 DefaultTestModel
 	defaultModel := strings.ToLower(openai.DefaultTestModel)
 	if pricing, ok := s.pricingData[defaultModel]; ok {

backend/internal/service/pricing_service_test.go

@@ -51,3 +51,20 @@ func TestGetModelPricing_OpenAIFallbackMatchedLoggedAsInfo(t *testing.T) {
 	require.True(t, logSink.ContainsMessageAtLevel("[Pricing] OpenAI fallback matched gpt-5.3-codex -> gpt-5.2-codex", "info"))
 	require.False(t, logSink.ContainsMessageAtLevel("[Pricing] OpenAI fallback matched gpt-5.3-codex -> gpt-5.2-codex", "warn"))
 }
+
+func TestGetModelPricing_Gpt54UsesStaticFallbackWhenRemoteMissing(t *testing.T) {
+	svc := &PricingService{
+		pricingData: map[string]*LiteLLMModelPricing{
+			"gpt-5.1-codex": &LiteLLMModelPricing{InputCostPerToken: 1.25e-6},
+		},
+	}
+
+	got := svc.GetModelPricing("gpt-5.4")
+	require.NotNil(t, got)
+	require.InDelta(t, 2.5e-6, got.InputCostPerToken, 1e-12)
+	require.InDelta(t, 1.5e-5, got.OutputCostPerToken, 1e-12)
+	require.InDelta(t, 2.5e-7, got.CacheReadInputTokenCost, 1e-12)
+	require.Equal(t, 272000, got.LongContextInputTokenThreshold)
+	require.InDelta(t, 2.0, got.LongContextInputCostMultiplier, 1e-12)
+	require.InDelta(t, 1.5, got.LongContextOutputCostMultiplier, 1e-12)
+}

backend/resources/model-pricing/model_prices_and_context_window.json

@@ -5140,6 +5140,39 @@
     "supports_vision": true,
     "supports_web_search": true
   },
+  "gpt-5.4": {
+    "cache_read_input_token_cost": 2.5e-07,
+    "input_cost_per_token": 2.5e-06,
+    "litellm_provider": "openai",
+    "max_input_tokens": 1050000,
+    "max_output_tokens": 128000,
+    "max_tokens": 128000,
+    "mode": "chat",
+    "output_cost_per_token": 1.5e-05,
+    "supported_endpoints": [
+      "/v1/chat/completions",
+      "/v1/responses"
+    ],
+    "supported_modalities": [
+      "text",
+      "image"
+    ],
+    "supported_output_modalities": [
+      "text",
+      "image"
+    ],
+    "supports_function_calling": true,
+    "supports_native_streaming": true,
+    "supports_parallel_function_calling": true,
+    "supports_pdf_input": true,
+    "supports_prompt_caching": true,
+    "supports_reasoning": true,
+    "supports_response_schema": true,
+    "supports_service_tier": true,
+    "supports_system_messages": true,
+    "supports_tool_choice": true,
+    "supports_vision": true
+  },
   "gpt-5.3-codex": {
     "cache_read_input_token_cost": 1.75e-07,
     "cache_read_input_token_cost_priority": 3.5e-07,

frontend/src/components/account/BulkEditAccountModal.vue

@@ -951,6 +951,7 @@ const allModels = [
   { value: 'claude-3-haiku-20240307', label: 'Claude 3 Haiku' },
   { value: 'gpt-5.3-codex', label: 'GPT-5.3 Codex' },
   { value: 'gpt-5.3-codex-spark', label: 'GPT-5.3 Codex Spark' },
+  { value: 'gpt-5.4', label: 'GPT-5.4' },
   { value: 'gpt-5.2-2025-12-11', label: 'GPT-5.2' },
   { value: 'gpt-5.2-codex', label: 'GPT-5.2 Codex' },
   { value: 'gpt-5.1-codex-max', label: 'GPT-5.1 Codex Max' },
@@ -1065,6 +1066,12 @@ const presetMappings = [
     to: 'gpt-5.3-codex-spark',
     color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400'
   },
+  {
+    label: 'GPT-5.4',
+    from: 'gpt-5.4',
+    to: 'gpt-5.4',
+    color: 'bg-rose-100 text-rose-700 hover:bg-rose-200 dark:bg-rose-900/30 dark:text-rose-400'
+  },
   {
     label: '5.2→5.3',
     from: 'gpt-5.2-codex',

frontend/src/components/keys/UseKeyModal.vue

@@ -670,6 +670,22 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
         xhigh: {}
       }
     },
+    'gpt-5.4': {
+      name: 'GPT-5.4',
+      limit: {
+        context: 1050000,
+        output: 128000
+      },
+      options: {
+        store: false
+      },
+      variants: {
+        low: {},
+        medium: {},
+        high: {},
+        xhigh: {}
+      }
+    },
     'gpt-5.3-codex-spark': {
       name: 'GPT-5.3 Codex Spark',
       limit: {

frontend/src/composables/__tests__/useModelWhitelist.spec.ts

@@ -2,6 +2,13 @@ import { describe, expect, it } from 'vitest'
 import { buildModelMappingObject, getModelsByPlatform } from '../useModelWhitelist'
 
 describe('useModelWhitelist', () => {
+  it('openai 模型列表包含 GPT-5.4 官方快照', () => {
+    const models = getModelsByPlatform('openai')
+
+    expect(models).toContain('gpt-5.4')
+    expect(models).toContain('gpt-5.4-2026-03-05')
+  })
+
   it('antigravity 模型列表包含图片模型兼容项', () => {
     const models = getModelsByPlatform('antigravity')
 
@@ -15,4 +22,12 @@ describe('useModelWhitelist', () => {
       'gemini-3.1-flash-image': 'gemini-3.1-flash-image'
     })
   })
+
+  it('whitelist 模式会保留 GPT-5.4 官方快照的精确映射', () => {
+    const mapping = buildModelMappingObject('whitelist', ['gpt-5.4-2026-03-05'], [])
+
+    expect(mapping).toEqual({
+      'gpt-5.4-2026-03-05': 'gpt-5.4-2026-03-05'
+    })
+  })
 })

frontend/src/composables/useModelWhitelist.ts

@@ -24,6 +24,8 @@ const openaiModels = [
   // GPT-5.2 系列
   'gpt-5.2', 'gpt-5.2-2025-12-11', 'gpt-5.2-chat-latest',
   'gpt-5.2-codex', 'gpt-5.2-pro', 'gpt-5.2-pro-2025-12-11',
+  // GPT-5.4 系列
+  'gpt-5.4', 'gpt-5.4-2026-03-05',
   // GPT-5.3 系列
   'gpt-5.3-codex', 'gpt-5.3-codex-spark',
   'chatgpt-4o-latest',
@@ -277,6 +279,7 @@ const openaiPresetMappings = [
   { label: 'GPT-5.3 Codex Spark', from: 'gpt-5.3-codex-spark', to: 'gpt-5.3-codex-spark', color: 'bg-teal-100 text-teal-700 hover:bg-teal-200 dark:bg-teal-900/30 dark:text-teal-400' },
   { label: 'GPT-5.1', from: 'gpt-5.1', to: 'gpt-5.1', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
   { label: 'GPT-5.2', from: 'gpt-5.2', to: 'gpt-5.2', color: 'bg-red-100 text-red-700 hover:bg-red-200 dark:bg-red-900/30 dark:text-red-400' },
+  { label: 'GPT-5.4', from: 'gpt-5.4', to: 'gpt-5.4', color: 'bg-rose-100 text-rose-700 hover:bg-rose-200 dark:bg-rose-900/30 dark:text-rose-400' },
   { label: 'GPT-5.1 Codex', from: 'gpt-5.1-codex', to: 'gpt-5.1-codex', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' }
 ]