oadmin/sub2api
1
0
Fork 0
Code Issues Pull Requests 1 Actions 1 Packages Projects Releases Wiki Activity

Merge pull request #1162 from remxcode/main

Browse Source 
feat(openai): 增加 gpt-5.4-mini/nano 模型支持与定价配置
This commit is contained in:
Wesley Liddick
2026-03-20 13:57:47 +08:00
committed by GitHub
parent 553a486d17 578608d301
commit a225a241d7
12 changed files with 310 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
backend/internal/service/billing_service.go
View File

@@ -221,6 +221,18 @@ func (s *BillingService) initFallbackPricing() {
LongContextInputMultiplier: openAIGPT54LongContextInputMultiplier,
LongContextOutputMultiplier: openAIGPT54LongContextOutputMultiplier,
}
s.fallbackPrices["gpt-5.4-mini"] = &ModelPricing{
InputPricePerToken: 7.5e-7,
OutputPricePerToken: 4.5e-6,
CacheReadPricePerToken: 7.5e-8,
SupportsCacheBreakdown: false,
}
s.fallbackPrices["gpt-5.4-nano"] = &ModelPricing{
InputPricePerToken: 2e-7,
OutputPricePerToken: 1.25e-6,
CacheReadPricePerToken: 2e-8,
SupportsCacheBreakdown: false,
}
// OpenAI GPT-5.2（本地兜底）
s.fallbackPrices["gpt-5.2"] = &ModelPricing{
InputPricePerToken: 1.75e-6,
@@ -294,6 +306,10 @@ func (s *BillingService) getFallbackPricing(model string) *ModelPricing {
if strings.Contains(modelLower, "gpt-5") || strings.Contains(modelLower, "codex") {
normalized := normalizeCodexModel(modelLower)
switch normalized {
case "gpt-5.4-mini":
return s.fallbackPrices["gpt-5.4-mini"]
case "gpt-5.4-nano":
return s.fallbackPrices["gpt-5.4-nano"]
case "gpt-5.4":
return s.fallbackPrices["gpt-5.4"]
case "gpt-5.2":

60
backend/internal/service/billing_service_test.go
View File

@@ -174,6 +174,30 @@ func TestGetModelPricing_OpenAIGPT54Fallback(t *testing.T) {
require.InDelta(t, 1.5, pricing.LongContextOutputMultiplier, 1e-12)
}
func TestGetModelPricing_OpenAIGPT54MiniFallback(t *testing.T) {
svc := newTestBillingService()
pricing, err := svc.GetModelPricing("gpt-5.4-mini")
require.NoError(t, err)
require.NotNil(t, pricing)
require.InDelta(t, 7.5e-7, pricing.InputPricePerToken, 1e-12)
require.InDelta(t, 4.5e-6, pricing.OutputPricePerToken, 1e-12)
require.InDelta(t, 7.5e-8, pricing.CacheReadPricePerToken, 1e-12)
require.Zero(t, pricing.LongContextInputThreshold)
}
func TestGetModelPricing_OpenAIGPT54NanoFallback(t *testing.T) {
svc := newTestBillingService()
pricing, err := svc.GetModelPricing("gpt-5.4-nano")
require.NoError(t, err)
require.NotNil(t, pricing)
require.InDelta(t, 2e-7, pricing.InputPricePerToken, 1e-12)
require.InDelta(t, 1.25e-6, pricing.OutputPricePerToken, 1e-12)
require.InDelta(t, 2e-8, pricing.CacheReadPricePerToken, 1e-12)
require.Zero(t, pricing.LongContextInputThreshold)
}
func TestCalculateCost_OpenAIGPT54LongContextAppliesWholeSessionMultipliers(t *testing.T) {
svc := newTestBillingService()
@@ -210,6 +234,8 @@ func TestGetFallbackPricing_FamilyMatching(t *testing.T) {
{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 gpt5.4 mini", model: "gpt-5.4-mini", expectedInput: 7.5e-7},
{name: "openai gpt5.4 nano", model: "gpt-5.4-nano", expectedInput: 2e-7},
{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},
@@ -564,6 +590,40 @@ func TestCalculateCostWithServiceTier_FlexAppliesHalfMultiplier(t *testing.T) {
require.InDelta(t, baseCost.TotalCost*0.5, flexCost.TotalCost, 1e-10)
}
func TestCalculateCostWithServiceTier_Gpt54MiniPriorityFallsBackToTierMultiplier(t *testing.T) {
svc := newTestBillingService()
tokens := UsageTokens{InputTokens: 120, OutputTokens: 30, CacheCreationTokens: 12, CacheReadTokens: 8}
baseCost, err := svc.CalculateCost("gpt-5.4-mini", tokens, 1.0)
require.NoError(t, err)
priorityCost, err := svc.CalculateCostWithServiceTier("gpt-5.4-mini", tokens, 1.0, "priority")
require.NoError(t, err)
require.InDelta(t, baseCost.InputCost*2, priorityCost.InputCost, 1e-10)
require.InDelta(t, baseCost.OutputCost*2, priorityCost.OutputCost, 1e-10)
require.InDelta(t, baseCost.CacheCreationCost*2, priorityCost.CacheCreationCost, 1e-10)
require.InDelta(t, baseCost.CacheReadCost*2, priorityCost.CacheReadCost, 1e-10)
require.InDelta(t, baseCost.TotalCost*2, priorityCost.TotalCost, 1e-10)
}
func TestCalculateCostWithServiceTier_Gpt54NanoFlexAppliesHalfMultiplier(t *testing.T) {
svc := newTestBillingService()
tokens := UsageTokens{InputTokens: 100, OutputTokens: 50, CacheCreationTokens: 40, CacheReadTokens: 20}
baseCost, err := svc.CalculateCost("gpt-5.4-nano", tokens, 1.0)
require.NoError(t, err)
flexCost, err := svc.CalculateCostWithServiceTier("gpt-5.4-nano", tokens, 1.0, "flex")
require.NoError(t, err)
require.InDelta(t, baseCost.InputCost*0.5, flexCost.InputCost, 1e-10)
require.InDelta(t, baseCost.OutputCost*0.5, flexCost.OutputCost, 1e-10)
require.InDelta(t, baseCost.CacheCreationCost*0.5, flexCost.CacheCreationCost, 1e-10)
require.InDelta(t, baseCost.CacheReadCost*0.5, flexCost.CacheReadCost, 1e-10)
require.InDelta(t, baseCost.TotalCost*0.5, flexCost.TotalCost, 1e-10)
}
func TestCalculateCostWithServiceTier_PriorityFallsBackToTierMultiplierWithoutExplicitPriorityPrice(t *testing.T) {
svc := newTestBillingService()
tokens := UsageTokens{InputTokens: 120, OutputTokens: 30, CacheCreationTokens: 12, CacheReadTokens: 8}

8
backend/internal/service/openai_codex_transform.go
View File

@@ -7,6 +7,8 @@ import (
var codexModelMap = map[string]string{
"gpt-5.4": "gpt-5.4",
"gpt-5.4-mini": "gpt-5.4-mini",
"gpt-5.4-nano": "gpt-5.4-nano",
"gpt-5.4-none": "gpt-5.4",
"gpt-5.4-low": "gpt-5.4",
"gpt-5.4-medium": "gpt-5.4",
@@ -225,6 +227,12 @@ func normalizeCodexModel(model string) string {
normalized := strings.ToLower(modelID)
if strings.Contains(normalized, "gpt-5.4-mini") || strings.Contains(normalized, "gpt 5.4 mini") {
return "gpt-5.4-mini"
}
if strings.Contains(normalized, "gpt-5.4-nano") || strings.Contains(normalized, "gpt 5.4 nano") {
return "gpt-5.4-nano"
}
if strings.Contains(normalized, "gpt-5.4") || strings.Contains(normalized, "gpt 5.4") {
return "gpt-5.4"
}

4
backend/internal/service/openai_codex_transform_test.go
View File

@@ -238,6 +238,10 @@ func TestNormalizeCodexModel_Gpt53(t *testing.T) {
"gpt-5.4-high": "gpt-5.4",
"gpt-5.4-chat-latest": "gpt-5.4",
"gpt 5.4": "gpt-5.4",
"gpt-5.4-mini": "gpt-5.4-mini",
"gpt 5.4 mini": "gpt-5.4-mini",
"gpt-5.4-nano": "gpt-5.4-nano",
"gpt 5.4 nano": "gpt-5.4-nano",
"gpt-5.3": "gpt-5.3-codex",
"gpt-5.3-codex": "gpt-5.3-codex",
"gpt-5.3-codex-xhigh": "gpt-5.3-codex",

28
backend/internal/service/pricing_service.go
View File

@@ -34,6 +34,22 @@ var (
Mode: "chat",
SupportsPromptCaching: true,
}
openAIGPT54MiniFallbackPricing = &LiteLLMModelPricing{
InputCostPerToken: 7.5e-07,
OutputCostPerToken: 4.5e-06,
CacheReadInputTokenCost: 7.5e-08,
LiteLLMProvider: "openai",
Mode: "chat",
SupportsPromptCaching: true,
}
openAIGPT54NanoFallbackPricing = &LiteLLMModelPricing{
InputCostPerToken: 2e-07,
OutputCostPerToken: 1.25e-06,
CacheReadInputTokenCost: 2e-08,
LiteLLMProvider: "openai",
Mode: "chat",
SupportsPromptCaching: true,
}
)
// LiteLLMModelPricing LiteLLM价格数据结构
@@ -723,6 +739,18 @@ func (s *PricingService) matchOpenAIModel(model string) *LiteLLMModelPricing {
}
}
if strings.HasPrefix(model, "gpt-5.4-mini") {
logger.With(zap.String("component", "service.pricing")).
Info(fmt.Sprintf("[Pricing] OpenAI fallback matched %s -> %s", model, "gpt-5.4-mini(static)"))
return openAIGPT54MiniFallbackPricing
}
if strings.HasPrefix(model, "gpt-5.4-nano") {
logger.With(zap.String("component", "service.pricing")).
Info(fmt.Sprintf("[Pricing] OpenAI fallback matched %s -> %s", model, "gpt-5.4-nano(static)"))
return openAIGPT54NanoFallbackPricing
}
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)"))

30
backend/internal/service/pricing_service_test.go
View File

@@ -98,6 +98,36 @@ func TestGetModelPricing_Gpt54UsesStaticFallbackWhenRemoteMissing(t *testing.T)
require.InDelta(t, 1.5, got.LongContextOutputCostMultiplier, 1e-12)
}
func TestGetModelPricing_Gpt54MiniUsesDedicatedStaticFallbackWhenRemoteMissing(t *testing.T) {
svc := &PricingService{
pricingData: map[string]*LiteLLMModelPricing{
"gpt-5.1-codex": {InputCostPerToken: 1.25e-6},
},
}
got := svc.GetModelPricing("gpt-5.4-mini")
require.NotNil(t, got)
require.InDelta(t, 7.5e-7, got.InputCostPerToken, 1e-12)
require.InDelta(t, 4.5e-6, got.OutputCostPerToken, 1e-12)
require.InDelta(t, 7.5e-8, got.CacheReadInputTokenCost, 1e-12)
require.Zero(t, got.LongContextInputTokenThreshold)
}
func TestGetModelPricing_Gpt54NanoUsesDedicatedStaticFallbackWhenRemoteMissing(t *testing.T) {
svc := &PricingService{
pricingData: map[string]*LiteLLMModelPricing{
"gpt-5.1-codex": {InputCostPerToken: 1.25e-6},
},
}
got := svc.GetModelPricing("gpt-5.4-nano")
require.NotNil(t, got)
require.InDelta(t, 2e-7, got.InputCostPerToken, 1e-12)
require.InDelta(t, 1.25e-6, got.OutputCostPerToken, 1e-12)
require.InDelta(t, 2e-8, got.CacheReadInputTokenCost, 1e-12)
require.Zero(t, got.LongContextInputTokenThreshold)
}
func TestParsePricingData_PreservesPriorityAndServiceTierFields(t *testing.T) {
raw := map[string]any{
"gpt-5.4": map[string]any{