diff --git a/backend/internal/pkg/apicompat/chatcompletions_responses_test.go b/backend/internal/pkg/apicompat/chatcompletions_responses_test.go
index d63ce664a7..4075f57791 100644
--- a/backend/internal/pkg/apicompat/chatcompletions_responses_test.go
+++ b/backend/internal/pkg/apicompat/chatcompletions_responses_test.go
@@ -53,6 +53,26 @@ func TestUsageConversionsPreserveCacheWriteTokens(t *testing.T) {
require.Equal(t, 200, roundTrip.InputTokensDetails.CacheWriteTokens)
}
+func TestResponsesUsageNestedCacheWritePresenceOverridesTopLevelAlias(t *testing.T) {
+ tests := []struct {
+ name string
+ nestedJSON string
+ want int
+ }{
+ {name: "explicit zero", nestedJSON: `{"cache_write_tokens":0}`, want: 0},
+ {name: "nonzero", nestedJSON: `{"cache_write_tokens":7}`, want: 7},
+ }
+
+ for _, tt := range tests {
+ t.Run(tt.name, func(t *testing.T) {
+ var usage ResponsesUsage
+ payload := []byte(`{"input_tokens":20,"output_tokens":2,"cache_creation_input_tokens":19,"input_tokens_details":` + tt.nestedJSON + `}`)
+ require.NoError(t, json.Unmarshal(payload, &usage))
+ require.Equal(t, tt.want, usage.CacheCreationInputTokens)
+ })
+ }
+}
+
func TestChatCompletionsToResponses_SystemMessage(t *testing.T) {
req := &ChatCompletionsRequest{
Model: "gpt-4o",
diff --git a/backend/internal/pkg/apicompat/types.go b/backend/internal/pkg/apicompat/types.go
index c020be94f8..8d96a1d3d3 100644
--- a/backend/internal/pkg/apicompat/types.go
+++ b/backend/internal/pkg/apicompat/types.go
@@ -332,6 +332,10 @@ type ResponsesUsage struct {
func (u *ResponsesUsage) UnmarshalJSON(data []byte) error {
type responsesUsageAlias ResponsesUsage
+ type cacheTokenPresence struct {
+ CacheCreationTokens *int `json:"cache_creation_tokens"`
+ CacheWriteTokens *int `json:"cache_write_tokens"`
+ }
var aux struct {
responsesUsageAlias
PromptTokens int `json:"prompt_tokens"`
@@ -345,6 +349,13 @@ func (u *ResponsesUsage) UnmarshalJSON(data []byte) error {
if err := json.Unmarshal(data, &aux); err != nil {
return err
}
+ var nestedPresence struct {
+ InputTokensDetails *cacheTokenPresence `json:"input_tokens_details"`
+ PromptTokensDetails *cacheTokenPresence `json:"prompt_tokens_details"`
+ }
+ if err := json.Unmarshal(data, &nestedPresence); err != nil {
+ return err
+ }
*u = ResponsesUsage(aux.responsesUsageAlias)
if u.InputTokens == 0 && aux.PromptTokens != 0 {
u.InputTokens = aux.PromptTokens
@@ -368,6 +379,20 @@ func (u *ResponsesUsage) UnmarshalJSON(data []byte) error {
if u.OutputTokensDetails == nil && aux.CompletionTokensDetails != nil {
u.OutputTokensDetails = aux.CompletionTokensDetails
}
+ var canonicalCacheCreationTokens *int
+ switch {
+ case nestedPresence.InputTokensDetails != nil && nestedPresence.InputTokensDetails.CacheWriteTokens != nil:
+ canonicalCacheCreationTokens = nestedPresence.InputTokensDetails.CacheWriteTokens
+ case nestedPresence.PromptTokensDetails != nil && nestedPresence.PromptTokensDetails.CacheWriteTokens != nil:
+ canonicalCacheCreationTokens = nestedPresence.PromptTokensDetails.CacheWriteTokens
+ case nestedPresence.InputTokensDetails != nil && nestedPresence.InputTokensDetails.CacheCreationTokens != nil:
+ canonicalCacheCreationTokens = nestedPresence.InputTokensDetails.CacheCreationTokens
+ case nestedPresence.PromptTokensDetails != nil && nestedPresence.PromptTokensDetails.CacheCreationTokens != nil:
+ canonicalCacheCreationTokens = nestedPresence.PromptTokensDetails.CacheCreationTokens
+ }
+ if canonicalCacheCreationTokens != nil {
+ u.CacheCreationInputTokens = max(*canonicalCacheCreationTokens, 0)
+ }
if u.TotalTokens == 0 && (u.InputTokens != 0 || u.OutputTokens != 0) {
u.TotalTokens = u.InputTokens + u.OutputTokens
}
diff --git a/backend/internal/pkg/openai/constants.go b/backend/internal/pkg/openai/constants.go
index c9d391df4e..79d799d774 100644
--- a/backend/internal/pkg/openai/constants.go
+++ b/backend/internal/pkg/openai/constants.go
@@ -18,6 +18,7 @@ type Model struct {
// DefaultModels OpenAI models list
var DefaultModels = []Model{
+ {ID: "gpt-5.6", Object: "model", Created: 1780876800, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.6 (Sol)"},
{ID: "gpt-5.6-sol", Object: "model", Created: 1780876800, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.6 Sol"},
{ID: "gpt-5.6-terra", Object: "model", Created: 1780876800, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.6 Terra"},
{ID: "gpt-5.6-luna", Object: "model", Created: 1780876800, OwnedBy: "openai", Type: "model", DisplayName: "GPT-5.6 Luna"},
diff --git a/backend/internal/pkg/openai/constants_test.go b/backend/internal/pkg/openai/constants_test.go
new file mode 100644
index 0000000000..f1f59b70f2
--- /dev/null
+++ b/backend/internal/pkg/openai/constants_test.go
@@ -0,0 +1,11 @@
+package openai
+
+import (
+ "testing"
+
+ "github.com/stretchr/testify/require"
+)
+
+func TestDefaultModelsIncludeBareGPT56Alias(t *testing.T) {
+ require.Contains(t, DefaultModelIDs(), "gpt-5.6")
+}
diff --git a/backend/internal/repository/usage_log_repo.go b/backend/internal/repository/usage_log_repo.go
index 341bdff57d..4359edc2ca 100644
--- a/backend/internal/repository/usage_log_repo.go
+++ b/backend/internal/repository/usage_log_repo.go
@@ -199,16 +199,24 @@ func appendRequestTypeOrStreamQueryFilter(query string, args []any, requestType
// buildRequestTypeFilterCondition 在 request_type 过滤时兼容 legacy 字段，避免历史数据漏查。
func buildRequestTypeFilterCondition(startArgIndex int, requestType int16) (string, []any) {
+ return buildRequestTypeFilterConditionWithAlias(startArgIndex, requestType, "")
+}
+
+func buildRequestTypeFilterConditionWithAlias(startArgIndex int, requestType int16, alias string) (string, []any) {
normalized := service.RequestTypeFromInt16(requestType)
requestTypeArg := int16(normalized)
+ prefix := ""
+ if alias != "" {
+ prefix = alias + "."
+ }
switch normalized {
case service.RequestTypeSync:
- return fmt.Sprintf("(request_type = $%d OR (request_type = %d AND stream = FALSE AND openai_ws_mode = FALSE))", startArgIndex, int16(service.RequestTypeUnknown)), []any{requestTypeArg}
+ return fmt.Sprintf("(%srequest_type = $%d OR (%srequest_type = %d AND %sstream = FALSE AND %sopenai_ws_mode = FALSE))", prefix, startArgIndex, prefix, int16(service.RequestTypeUnknown), prefix, prefix), []any{requestTypeArg}
case service.RequestTypeStream:
- return fmt.Sprintf("(request_type = $%d OR (request_type = %d AND stream = TRUE AND openai_ws_mode = FALSE))", startArgIndex, int16(service.RequestTypeUnknown)), []any{requestTypeArg}
+ return fmt.Sprintf("(%srequest_type = $%d OR (%srequest_type = %d AND %sstream = TRUE AND %sopenai_ws_mode = FALSE))", prefix, startArgIndex, prefix, int16(service.RequestTypeUnknown), prefix, prefix), []any{requestTypeArg}
case service.RequestTypeWSV2:
- return fmt.Sprintf("(request_type = $%d OR (request_type = %d AND openai_ws_mode = TRUE))", startArgIndex, int16(service.RequestTypeUnknown)), []any{requestTypeArg}
+ return fmt.Sprintf("(%srequest_type = $%d OR (%srequest_type = %d AND %sopenai_ws_mode = TRUE))", prefix, startArgIndex, prefix, int16(service.RequestTypeUnknown), prefix), []any{requestTypeArg}
default:
- return fmt.Sprintf("request_type = $%d", startArgIndex), []any{requestTypeArg}
+ return fmt.Sprintf("%srequest_type = $%d", prefix, startArgIndex), []any{requestTypeArg}
}
}
diff --git a/backend/internal/repository/usage_log_repo_breakdown_test.go b/backend/internal/repository/usage_log_repo_breakdown_test.go
index da62e8dd56..1fb99b0d64 100644
--- a/backend/internal/repository/usage_log_repo_breakdown_test.go
+++ b/backend/internal/repository/usage_log_repo_breakdown_test.go
@@ -3,9 +3,14 @@
package repository
import (
+ "context"
+ "regexp"
"testing"
+ "time"
+ "github.com/DATA-DOG/go-sqlmock"
"github.com/Wei-Shaw/sub2api/internal/pkg/usagestats"
+ "github.com/Wei-Shaw/sub2api/internal/service"
"github.com/stretchr/testify/require"
)
@@ -48,3 +53,27 @@ func TestResolveModelDimensionExpression(t *testing.T) {
})
}
}
+
+func TestGetUserBreakdownStatsRequestTypeIncludesLegacyFallback(t *testing.T) {
+ db, mock := newSQLMock(t)
+ repo := &usageLogRepository{sql: db}
+ start := time.Date(2026, 7, 1, 0, 0, 0, 0, time.UTC)
+ end := start.Add(24 * time.Hour)
+ requestType := int16(service.RequestTypeStream)
+
+ legacyFilter := `(ul.request_type = $3 OR (ul.request_type = 0 AND ul.stream = TRUE AND ul.openai_ws_mode = FALSE))`
+ mock.ExpectQuery(regexp.QuoteMeta(legacyFilter)).
+ WithArgs(start, end, requestType).
+ WillReturnRows(sqlmock.NewRows([]string{
+ "user_id", "email", "requests", "input_tokens", "output_tokens",
+ "cache_tokens", "total_tokens", "cost", "actual_cost", "account_cost",
+ }))
+
+ rows, err := repo.GetUserBreakdownStats(context.Background(), start, end, usagestats.UserBreakdownDimension{
+ RequestType: &requestType,
+ }, 0)
+
+ require.NoError(t, err)
+ require.Empty(t, rows)
+ require.NoError(t, mock.ExpectationsWereMet())
+}
diff --git a/backend/internal/repository/usage_log_repo_trend.go b/backend/internal/repository/usage_log_repo_trend.go
index 7302296683..f2e0045ff5 100644
--- a/backend/internal/repository/usage_log_repo_trend.go
+++ b/backend/internal/repository/usage_log_repo_trend.go
@@ -642,8 +642,9 @@ func (r *usageLogRepository) GetUserBreakdownStats(ctx context.Context, startTim
args = append(args, dim.AccountID)
}
if dim.RequestType != nil {
- query += fmt.Sprintf(" AND ul.request_type = $%d", len(args)+1)
- args = append(args, *dim.RequestType)
+ condition, conditionArgs := buildRequestTypeFilterConditionWithAlias(len(args)+1, *dim.RequestType, "ul")
+ query += " AND " + condition
+ args = append(args, conditionArgs...)
}
if dim.Stream != nil {
query += fmt.Sprintf(" AND ul.stream = $%d", len(args)+1)
diff --git a/backend/internal/service/billing_service.go b/backend/internal/service/billing_service.go
index 71d74ae6e0..3dfd500b05 100644
--- a/backend/internal/service/billing_service.go
+++ b/backend/internal/service/billing_service.go
@@ -293,6 +293,9 @@ func (s *BillingService) initFallbackPricing() {
CacheCreationPricePerTokenPriority: 12.5e-6,
CacheReadPricePerToken: 0.5e-6,
CacheReadPricePerTokenPriority: 1e-6,
+ LongContextInputThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputMultiplier: openAIGPT54LongContextOutputMultiplier,
}
s.fallbackPrices["gpt-5.6-terra"] = &ModelPricing{
InputPricePerToken: 2.5e-6,
@@ -303,6 +306,9 @@ func (s *BillingService) initFallbackPricing() {
CacheCreationPricePerTokenPriority: 6.25e-6,
CacheReadPricePerToken: 0.25e-6,
CacheReadPricePerTokenPriority: 0.5e-6,
+ LongContextInputThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputMultiplier: openAIGPT54LongContextOutputMultiplier,
}
s.fallbackPrices["gpt-5.6-luna"] = &ModelPricing{
InputPricePerToken: 1e-6,
@@ -313,6 +319,9 @@ func (s *BillingService) initFallbackPricing() {
CacheCreationPricePerTokenPriority: 2.5e-6,
CacheReadPricePerToken: 0.1e-6,
CacheReadPricePerTokenPriority: 0.2e-6,
+ LongContextInputThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputMultiplier: openAIGPT54LongContextOutputMultiplier,
}
s.fallbackPrices["gpt-5.4-mini"] = &ModelPricing{
@@ -1100,7 +1109,7 @@ func (s *BillingService) applyModelSpecificPricingPolicy(model string, pricing *
if !isGPT56 && !usesLegacyLongContextPricing {
return pricing
}
- needsLongContextPolicy := usesLegacyLongContextPricing &&
+ needsLongContextPolicy := (isGPT56 || usesLegacyLongContextPricing) &&
(pricing.LongContextInputThreshold <= 0 || pricing.LongContextInputMultiplier <= 0 || pricing.LongContextOutputMultiplier <= 0)
needsCacheCreationPolicy := isGPT56 && !pricing.CacheCreationPriceExplicit && (pricing.CacheCreationPricePerToken <= 0 ||
(pricing.InputPricePerTokenPriority > 0 && pricing.CacheCreationPricePerTokenPriority <= 0))
@@ -1108,7 +1117,7 @@ func (s *BillingService) applyModelSpecificPricingPolicy(model string, pricing *
return pricing
}
cloned := *pricing
- if isGPT56 {
+ if isGPT56 && !cloned.CacheCreationPriceExplicit {
if cloned.CacheCreationPricePerToken <= 0 {
cloned.CacheCreationPricePerToken = cloned.InputPricePerToken * 1.25
}
@@ -1116,7 +1125,7 @@ func (s *BillingService) applyModelSpecificPricingPolicy(model string, pricing *
cloned.CacheCreationPricePerTokenPriority = cloned.InputPricePerTokenPriority * 1.25
}
}
- if usesLegacyLongContextPricing {
+ if isGPT56 || usesLegacyLongContextPricing {
if cloned.LongContextInputThreshold <= 0 {
cloned.LongContextInputThreshold = openAIGPT54LongContextInputThreshold
}
diff --git a/backend/internal/service/openai_gateway_chat_completions_raw.go b/backend/internal/service/openai_gateway_chat_completions_raw.go
index 28c95f7ff8..6def67c7ec 100644
--- a/backend/internal/service/openai_gateway_chat_completions_raw.go
+++ b/backend/internal/service/openai_gateway_chat_completions_raw.go
@@ -2,14 +2,12 @@ package service
import (
"context"
- "encoding/json"
"errors"
"fmt"
"net/http"
"strings"
"time"
- "github.com/Wei-Shaw/sub2api/internal/pkg/apicompat"
"github.com/Wei-Shaw/sub2api/internal/pkg/logger"
"github.com/Wei-Shaw/sub2api/internal/pkg/xai"
"github.com/Wei-Shaw/sub2api/internal/util/responseheaders"
@@ -411,16 +409,9 @@ func (s *OpenAIGatewayService) bufferRawChatCompletions(
return nil, fmt.Errorf("read upstream body: %w", err)
}
- var ccResp apicompat.ChatCompletionsResponse
var usage OpenAIUsage
- if err := json.Unmarshal(respBody, &ccResp); err == nil && ccResp.Usage != nil {
- usage = OpenAIUsage{
- InputTokens: ccResp.Usage.PromptTokens,
- OutputTokens: ccResp.Usage.CompletionTokens,
- }
- if ccResp.Usage.PromptTokensDetails != nil {
- usage.CacheReadInputTokens = ccResp.Usage.PromptTokensDetails.CachedTokens
- }
+ if parsedUsage, ok := extractOpenAIUsageFromJSONBytes(respBody); ok {
+ usage = parsedUsage
}
if s.responseHeaderFilter != nil {
diff --git a/backend/internal/service/openai_gateway_chat_completions_raw_test.go b/backend/internal/service/openai_gateway_chat_completions_raw_test.go
index 5f4b5bcd2d..61663e636a 100644
--- a/backend/internal/service/openai_gateway_chat_completions_raw_test.go
+++ b/backend/internal/service/openai_gateway_chat_completions_raw_test.go
@@ -155,6 +155,57 @@ func TestForwardAsRawChatCompletions_PreservesMappedGPT56MaxEffort(t *testing.T)
require.Equal(t, "max", *result.ReasoningEffort)
}
+func TestForwardAsRawChatCompletions_NonStreamingCapturesCacheWriteUsage(t *testing.T) {
+ gin.SetMode(gin.TestMode)
+
+ tests := []struct {
+ name string
+ usageJSON string
+ wantWrite int
+ }{
+ {
+ name: "positive cache write",
+ usageJSON: `{"prompt_tokens":12,"completion_tokens":3,"total_tokens":15,"prompt_tokens_details":{"cached_tokens":4,"cache_write_tokens":6}}`,
+ wantWrite: 6,
+ },
+ {
+ name: "nested zero overrides legacy alias",
+ usageJSON: `{"prompt_tokens":12,"completion_tokens":3,"total_tokens":15,"cache_creation_input_tokens":19,"prompt_tokens_details":{"cached_tokens":4,"cache_write_tokens":0}}`,
+ wantWrite: 0,
+ },
+ }
+
+ for _, tt := range tests {
+ t.Run(tt.name, func(t *testing.T) {
+ body := []byte(`{"model":"gpt-5.6","messages":[{"role":"user","content":"hello"}],"stream":false}`)
+ rec := httptest.NewRecorder()
+ c, _ := gin.CreateTestContext(rec)
+ c.Request = httptest.NewRequest(http.MethodPost, "/v1/chat/completions", bytes.NewReader(body))
+ c.Request.Header.Set("Content-Type", "application/json")
+
+ upstream := &httpUpstreamRecorder{resp: &http.Response{
+ StatusCode: http.StatusOK,
+ Header: http.Header{"Content-Type": []string{"application/json"}},
+ Body: io.NopCloser(strings.NewReader(
+ `{"id":"chatcmpl_cache","object":"chat.completion","model":"gpt-5.6","choices":[{"index":0,"message":{"role":"assistant","content":"ok"},"finish_reason":"stop"}],"usage":` + tt.usageJSON + `}`,
+ )),
+ }}
+ svc := &OpenAIGatewayService{
+ cfg: rawChatCompletionsTestConfig(),
+ httpUpstream: upstream,
+ }
+
+ result, err := svc.forwardAsRawChatCompletions(context.Background(), c, rawChatCompletionsTestAccount(), body, "")
+
+ require.NoError(t, err)
+ require.NotNil(t, result)
+ require.Equal(t, 12, result.Usage.InputTokens)
+ require.Equal(t, 4, result.Usage.CacheReadInputTokens)
+ require.Equal(t, tt.wantWrite, result.Usage.CacheCreationInputTokens)
+ })
+ }
+}
+
func TestForwardAsRawChatCompletions_PreservesDeepSeekReasoningContentNonStreaming(t *testing.T) {
gin.SetMode(gin.TestMode)
diff --git a/backend/internal/service/openai_gateway_messages.go b/backend/internal/service/openai_gateway_messages.go
index 52ec34f27e..b1e99f8da9 100644
--- a/backend/internal/service/openai_gateway_messages.go
+++ b/backend/internal/service/openai_gateway_messages.go
@@ -1120,13 +1120,6 @@ func copyOpenAIUsageFromResponsesUsage(usage *apicompat.ResponsesUsage) OpenAIUs
}
if usage.InputTokensDetails != nil {
result.CacheReadInputTokens = usage.InputTokensDetails.CachedTokens
- if result.CacheCreationInputTokens == 0 {
- if usage.InputTokensDetails.CacheWriteTokens > 0 {
- result.CacheCreationInputTokens = usage.InputTokensDetails.CacheWriteTokens
- } else {
- result.CacheCreationInputTokens = usage.InputTokensDetails.CacheCreationTokens
- }
- }
}
return result
}
diff --git a/backend/internal/service/openai_gateway_messages_usage_test.go b/backend/internal/service/openai_gateway_messages_usage_test.go
new file mode 100644
index 0000000000..f884e74595
--- /dev/null
+++ b/backend/internal/service/openai_gateway_messages_usage_test.go
@@ -0,0 +1,28 @@
+//go:build unit
+
+package service
+
+import (
+ "testing"
+
+ "github.com/Wei-Shaw/sub2api/internal/pkg/apicompat"
+ "github.com/stretchr/testify/require"
+)
+
+func TestCopyOpenAIUsageFromResponsesUsageTrustsCanonicalCacheCreationValue(t *testing.T) {
+ usage := &apicompat.ResponsesUsage{
+ InputTokens: 20,
+ OutputTokens: 2,
+ CacheCreationInputTokens: 0,
+ InputTokensDetails: &apicompat.ResponsesInputTokensDetails{
+ CachedTokens: 3,
+ CacheWriteTokens: 19,
+ },
+ }
+
+ got := copyOpenAIUsageFromResponsesUsage(usage)
+
+ require.Equal(t, 20, got.InputTokens)
+ require.Equal(t, 3, got.CacheReadInputTokens)
+ require.Zero(t, got.CacheCreationInputTokens)
+}
diff --git a/backend/internal/service/openai_gateway_response_handling.go b/backend/internal/service/openai_gateway_response_handling.go
index 313a255341..5841ce6556 100644
--- a/backend/internal/service/openai_gateway_response_handling.go
+++ b/backend/internal/service/openai_gateway_response_handling.go
@@ -772,9 +772,16 @@ func openAIUsageFromGJSON(value gjson.Result) (OpenAIUsage, bool) {
}
func openAICacheReadTokensFromUsage(value gjson.Result) int {
- return firstPositiveGJSONInt(
+ for _, nested := range []gjson.Result{
value.Get("input_tokens_details.cached_tokens"),
value.Get("prompt_tokens_details.cached_tokens"),
+ } {
+ if nested.Exists() {
+ return max(int(nested.Int()), 0)
+ }
+ }
+
+ return firstPositiveGJSONInt(
value.Get("cache_read_input_tokens"),
value.Get("cache_read_tokens"),
value.Get("cached_tokens"),
@@ -782,11 +789,18 @@ func openAICacheReadTokensFromUsage(value gjson.Result) int {
}
func openAICacheCreationTokensFromUsage(value gjson.Result) int {
- return firstPositiveGJSONInt(
+ for _, nested := range []gjson.Result{
value.Get("input_tokens_details.cache_write_tokens"),
value.Get("prompt_tokens_details.cache_write_tokens"),
value.Get("input_tokens_details.cache_creation_tokens"),
value.Get("prompt_tokens_details.cache_creation_tokens"),
+ } {
+ if nested.Exists() {
+ return max(int(nested.Int()), 0)
+ }
+ }
+
+ return firstPositiveGJSONInt(
value.Get("cache_write_tokens"),
value.Get("cache_creation_input_tokens"),
value.Get("cache_write_input_tokens"),
diff --git a/backend/internal/service/openai_gateway_service_test.go b/backend/internal/service/openai_gateway_service_test.go
index 3fdef698db..b33918652f 100644
--- a/backend/internal/service/openai_gateway_service_test.go
+++ b/backend/internal/service/openai_gateway_service_test.go
@@ -2821,6 +2821,14 @@ func TestExtractOpenAIUsageFromJSONBytes_AcceptsResponseAndChatUsageShapes(t *te
usage, ok = extractOpenAIUsageFromJSONBytes([]byte(`{"usage":{"input_tokens":20,"output_tokens":2,"cache_creation_input_tokens":19,"input_tokens_details":{"cache_write_tokens":7}}}`))
require.True(t, ok)
require.Equal(t, 7, usage.CacheCreationInputTokens, "官方嵌套字段应优先于兼容顶层别名")
+
+ usage, ok = extractOpenAIUsageFromJSONBytes([]byte(`{"usage":{"input_tokens":20,"output_tokens":2,"cache_creation_input_tokens":19,"input_tokens_details":{"cache_write_tokens":0}}}`))
+ require.True(t, ok)
+ require.Zero(t, usage.CacheCreationInputTokens, "官方嵌套字段显式为零时仍应优先于兼容顶层别名")
+
+ usage, ok = extractOpenAIUsageFromJSONBytes([]byte(`{"usage":{"input_tokens":20,"output_tokens":2,"cache_read_input_tokens":19,"input_tokens_details":{"cached_tokens":0}}}`))
+ require.True(t, ok)
+ require.Zero(t, usage.CacheReadInputTokens, "官方嵌套缓存读取字段显式为零时仍应优先于兼容顶层别名")
}
func TestExtractCodexFinalResponse_SampleReplay(t *testing.T) {
diff --git a/backend/internal/service/openai_model_alias.go b/backend/internal/service/openai_model_alias.go
index 338de9e8a5..4ddf8110c3 100644
--- a/backend/internal/service/openai_model_alias.go
+++ b/backend/internal/service/openai_model_alias.go
@@ -71,6 +71,14 @@ func normalizeKnownOpenAICodexModel(model string) string {
return "gpt-5.6-terra"
case strings.Contains(normalized, "gpt-5.6-luna"):
return "gpt-5.6-luna"
+ case normalized == "gpt-5.6":
+ return "gpt-5.6-sol"
+ case strings.HasPrefix(normalized, "gpt-5.6-"):
+ suffix := strings.TrimPrefix(normalized, "gpt-5.6-")
+ if suffix == "max" || isKnownCodexModelSuffix(suffix) {
+ return "gpt-5.6-sol"
+ }
+ return ""
case strings.Contains(normalized, "gpt-5.5-pro"):
return "gpt-5.5-pro"
case strings.Contains(normalized, "gpt-5.5"):
@@ -102,6 +110,12 @@ func normalizeKnownOpenAICodexModel(model string) string {
// （含大小写/路径/后缀变体）或已归一化的基名，两者均能正确识别。
func isOpenAIGPT56Model(model string) bool {
normalized := canonicalizeOpenAIModelAliasSpelling(model)
+ if normalized == "gpt-5.6" {
+ return true
+ }
+ if suffix, ok := strings.CutPrefix(normalized, "gpt-5.6-"); ok && (suffix == "max" || isKnownCodexModelSuffix(suffix)) {
+ return true
+ }
for _, prefix := range []string{"gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"} {
if normalized == prefix || strings.HasPrefix(normalized, prefix+"-") {
return true
diff --git a/backend/internal/service/openai_model_alias_test.go b/backend/internal/service/openai_model_alias_test.go
new file mode 100644
index 0000000000..4cb27206c9
--- /dev/null
+++ b/backend/internal/service/openai_model_alias_test.go
@@ -0,0 +1,36 @@
+package service
+
+import (
+ "testing"
+
+ "github.com/stretchr/testify/require"
+)
+
+func TestNormalizeKnownOpenAICodexModel_BareGPT56RoutesToSol(t *testing.T) {
+ tests := map[string]string{
+ "gpt-5.6": "gpt-5.6-sol",
+ "openai/gpt-5.6": "gpt-5.6-sol",
+ "gpt5.6": "gpt-5.6-sol",
+ "gpt-5.6-high": "gpt-5.6-sol",
+ "gpt-5.6-max": "gpt-5.6-sol",
+ "gpt-5.6-2026-07-09": "gpt-5.6-sol",
+ "openai/gpt-5.6-max": "gpt-5.6-sol",
+ }
+
+ for input, expected := range tests {
+ t.Run(input, func(t *testing.T) {
+ require.Equal(t, expected, normalizeKnownOpenAICodexModel(input))
+ })
+ }
+}
+
+func TestUsageBillingModelCandidates_BareGPT56IncludesSol(t *testing.T) {
+ require.Equal(t,
+ []string{"gpt-5.6", "gpt-5.6-sol"},
+ usageBillingModelCandidates("gpt-5.6"),
+ )
+ require.Equal(t,
+ []string{"openai/gpt-5.6", "gpt-5.6", "gpt-5.6-sol"},
+ usageBillingModelCandidates("openai/gpt-5.6"),
+ )
+}
diff --git a/backend/internal/service/openai_model_mapping_test.go b/backend/internal/service/openai_model_mapping_test.go
index 0f65740be7..f2ceb3551c 100644
--- a/backend/internal/service/openai_model_mapping_test.go
+++ b/backend/internal/service/openai_model_mapping_test.go
@@ -252,6 +252,18 @@ func TestNormalizeOpenAIModelForUpstream(t *testing.T) {
model string
want string
}{
+ {
+ name: "oauth routes bare GPT-5.6 alias to Sol",
+ account: &Account{Type: AccountTypeOAuth},
+ model: "gpt-5.6",
+ want: "gpt-5.6-sol",
+ },
+ {
+ name: "oauth routes provider-prefixed GPT-5.6 alias to Sol",
+ account: &Account{Type: AccountTypeOAuth},
+ model: "openai/gpt-5.6",
+ want: "gpt-5.6-sol",
+ },
{
name: "oauth preserves unknown non codex model",
account: &Account{Type: AccountTypeOAuth},
@@ -282,6 +294,12 @@ func TestNormalizeOpenAIModelForUpstream(t *testing.T) {
model: "codex-auto-review",
want: "codex-auto-review",
},
+ {
+ name: "apikey preserves official bare GPT-5.6 alias",
+ account: &Account{Type: AccountTypeAPIKey},
+ model: "gpt-5.6",
+ want: "gpt-5.6",
+ },
{
name: "apikey preserves custom compatible model",
account: &Account{Type: AccountTypeAPIKey},
diff --git a/backend/internal/service/openai_ws_v2/passthrough_relay.go b/backend/internal/service/openai_ws_v2/passthrough_relay.go
index be19438eaa..d41abaac3a 100644
--- a/backend/internal/service/openai_ws_v2/passthrough_relay.go
+++ b/backend/internal/service/openai_ws_v2/passthrough_relay.go
@@ -816,6 +816,13 @@ func openAICacheCreationTokensFromUsage(value gjson.Result) int {
"prompt_tokens_details.cache_write_tokens",
"input_tokens_details.cache_creation_tokens",
"prompt_tokens_details.cache_creation_tokens",
+ } {
+ result := value.Get(field)
+ if result.Exists() {
+ return max(int(result.Int()), 0)
+ }
+ }
+ for _, field := range []string{
"cache_write_tokens",
"cache_creation_input_tokens",
"cache_write_input_tokens",
diff --git a/backend/internal/service/openai_ws_v2/passthrough_relay_internal_test.go b/backend/internal/service/openai_ws_v2/passthrough_relay_internal_test.go
index cb2bd9cc31..f92117a2bf 100644
--- a/backend/internal/service/openai_ws_v2/passthrough_relay_internal_test.go
+++ b/backend/internal/service/openai_ws_v2/passthrough_relay_internal_test.go
@@ -335,6 +335,13 @@ func TestParseUsageAndAccumulateAcceptsChatUsageAliases(t *testing.T) {
require.Equal(t, got, state.usage)
}
+func TestOpenAICacheCreationTokensFromUsageNestedZeroWins(t *testing.T) {
+ t.Parallel()
+
+ usage := gjson.Parse(`{"input_tokens_details":{"cache_write_tokens":0},"cache_creation_input_tokens":19}`)
+ require.Zero(t, openAICacheCreationTokensFromUsage(usage))
+}
+
func TestEmitTurnCompleteCoverage(t *testing.T) {
t.Parallel()
diff --git a/backend/internal/service/pricing_service.go b/backend/internal/service/pricing_service.go
index 4528a8886f..d0ae3c7d96 100644
--- a/backend/internal/service/pricing_service.go
+++ b/backend/internal/service/pricing_service.go
@@ -44,6 +44,9 @@ var (
CacheCreationInputTokenCostPriority: 1.25e-05,
CacheReadInputTokenCost: 5e-07,
CacheReadInputTokenCostPriority: 1e-06,
+ LongContextInputTokenThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputCostMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputCostMultiplier: openAIGPT54LongContextOutputMultiplier,
SupportsServiceTier: true,
LiteLLMProvider: "openai",
Mode: "chat",
@@ -58,6 +61,9 @@ var (
CacheCreationInputTokenCostPriority: 6.25e-06,
CacheReadInputTokenCost: 2.5e-07,
CacheReadInputTokenCostPriority: 5e-07,
+ LongContextInputTokenThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputCostMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputCostMultiplier: openAIGPT54LongContextOutputMultiplier,
SupportsServiceTier: true,
LiteLLMProvider: "openai",
Mode: "chat",
@@ -72,6 +78,9 @@ var (
CacheCreationInputTokenCostPriority: 2.5e-06,
CacheReadInputTokenCost: 1e-07,
CacheReadInputTokenCostPriority: 2e-07,
+ LongContextInputTokenThreshold: openAIGPT54LongContextInputThreshold,
+ LongContextInputCostMultiplier: openAIGPT54LongContextInputMultiplier,
+ LongContextOutputCostMultiplier: openAIGPT54LongContextOutputMultiplier,
SupportsServiceTier: true,
LiteLLMProvider: "openai",
Mode: "chat",
@@ -140,6 +149,9 @@ type LiteLLMRawEntry struct {
CacheCreationInputTokenCostAbove1hr *float64 `json:"cache_creation_input_token_cost_above_1hr"`
CacheReadInputTokenCost *float64 `json:"cache_read_input_token_cost"`
CacheReadInputTokenCostPriority *float64 `json:"cache_read_input_token_cost_priority"`
+ LongContextInputTokenThreshold *int `json:"long_context_input_token_threshold"`
+ LongContextInputCostMultiplier *float64 `json:"long_context_input_cost_multiplier"`
+ LongContextOutputCostMultiplier *float64 `json:"long_context_output_cost_multiplier"`
SupportsServiceTier bool `json:"supports_service_tier"`
LiteLLMProvider string `json:"litellm_provider"`
Mode string `json:"mode"`
@@ -462,6 +474,15 @@ func (s *PricingService) parsePricingData(body []byte) (map[string]*LiteLLMModel
if entry.CacheReadInputTokenCostPriority != nil {
pricing.CacheReadInputTokenCostPriority = *entry.CacheReadInputTokenCostPriority
}
+ if entry.LongContextInputTokenThreshold != nil {
+ pricing.LongContextInputTokenThreshold = *entry.LongContextInputTokenThreshold
+ }
+ if entry.LongContextInputCostMultiplier != nil {
+ pricing.LongContextInputCostMultiplier = *entry.LongContextInputCostMultiplier
+ }
+ if entry.LongContextOutputCostMultiplier != nil {
+ pricing.LongContextOutputCostMultiplier = *entry.LongContextOutputCostMultiplier
+ }
if entry.OutputCostPerImage != nil {
pricing.OutputCostPerImage = *entry.OutputCostPerImage
}
@@ -713,6 +734,12 @@ func normalizeModelNameForPricing(model string) string {
model = strings.TrimLeft(model, "/")
if canonical := canonicalizeOpenAIModelAliasSpelling(model); canonical != "" {
+ if canonical == "gpt-5.6" {
+ return "gpt-5.6-sol"
+ }
+ if suffix, ok := strings.CutPrefix(canonical, "gpt-5.6-"); ok && (suffix == "max" || isKnownCodexModelSuffix(suffix)) {
+ return "gpt-5.6-sol"
+ }
return canonical
}
return model
diff --git a/backend/internal/service/pricing_service_test.go b/backend/internal/service/pricing_service_test.go
index 19098df917..b68d034638 100644
--- a/backend/internal/service/pricing_service_test.go
+++ b/backend/internal/service/pricing_service_test.go
@@ -22,6 +22,9 @@ func TestParsePricingData_ParsesPriorityAndServiceTierFields(t *testing.T) {
"cache_creation_input_token_cost_priority": 0.000005,
"cache_read_input_token_cost": 0.00000025,
"cache_read_input_token_cost_priority": 0.0000005,
+ "long_context_input_token_threshold": 272000,
+ "long_context_input_cost_multiplier": 2,
+ "long_context_output_cost_multiplier": 1.5,
"supports_service_tier": true,
"supports_prompt_caching": true,
"litellm_provider": "openai",
@@ -37,6 +40,9 @@ func TestParsePricingData_ParsesPriorityAndServiceTierFields(t *testing.T) {
require.InDelta(t, 3e-5, pricing.OutputCostPerTokenPriority, 1e-12)
require.InDelta(t, 5e-6, pricing.CacheCreationInputTokenCostPriority, 1e-12)
require.InDelta(t, 5e-7, pricing.CacheReadInputTokenCostPriority, 1e-12)
+ require.Equal(t, 272000, pricing.LongContextInputTokenThreshold)
+ require.InDelta(t, 2.0, pricing.LongContextInputCostMultiplier, 1e-12)
+ require.InDelta(t, 1.5, pricing.LongContextOutputCostMultiplier, 1e-12)
require.True(t, pricing.SupportsServiceTier)
}
@@ -72,7 +78,9 @@ func TestBillingService_GPT56CacheWritePricingUsesOfficialMultiplier(t *testing.
require.NoError(t, err)
require.InDelta(t, tt.input*1.25, pricing.CacheCreationPricePerToken, 1e-12)
require.InDelta(t, tt.inputPriority*1.25, pricing.CacheCreationPricePerTokenPriority, 1e-12)
- require.Zero(t, pricing.LongContextInputThreshold)
+ require.Equal(t, 272000, pricing.LongContextInputThreshold)
+ require.InDelta(t, 2.0, pricing.LongContextInputMultiplier, 1e-12)
+ require.InDelta(t, 1.5, pricing.LongContextOutputMultiplier, 1e-12)
tokens := UsageTokens{InputTokens: 700, OutputTokens: 50, CacheCreationTokens: 200, CacheReadTokens: 100}
standard, err := svc.CalculateCostWithServiceTier(tt.model, tokens, 1, "")
@@ -90,25 +98,87 @@ func TestBillingService_GPT56CacheWritePricingUsesOfficialMultiplier(t *testing.
}
}
-func TestBillingService_GPT56DoesNotUseLegacyLongContextMultiplier(t *testing.T) {
- model := "gpt-5.6-sol"
- pricingSvc := &PricingService{pricingData: map[string]*LiteLLMModelPricing{
- model: {
- InputCostPerToken: 5e-6,
- OutputCostPerToken: 30e-6,
- CacheReadInputTokenCost: 0.5e-6,
- },
- }}
- svc := NewBillingService(&config.Config{}, pricingSvc)
- tokens := UsageTokens{InputTokens: 100000, CacheCreationTokens: 173000, OutputTokens: 10}
+func TestBillingService_GPT56UsesLongContextPricingAcrossModelsAndTiers(t *testing.T) {
+ models := []struct {
+ name string
+ input, cached float64
+ cacheWrite, output float64
+ }{
+ {name: "gpt-5.6-sol", input: 5e-6, cached: 0.5e-6, cacheWrite: 6.25e-6, output: 30e-6},
+ {name: "gpt-5.6-terra", input: 2.5e-6, cached: 0.25e-6, cacheWrite: 3.125e-6, output: 15e-6},
+ {name: "gpt-5.6-luna", input: 1e-6, cached: 0.1e-6, cacheWrite: 1.25e-6, output: 6e-6},
+ }
+ tiers := []struct {
+ name string
+ priceScale float64
+ }{
+ {name: "standard", priceScale: 1},
+ {name: "priority", priceScale: 2},
+ {name: "flex", priceScale: 0.5},
+ }
+ tokens := UsageTokens{
+ InputTokens: 100000,
+ CacheCreationTokens: 100000,
+ CacheReadTokens: 73000,
+ OutputTokens: 10,
+ }
- cost, err := svc.CalculateCost(model, tokens, 1)
+ for _, model := range models {
+ for _, tier := range tiers {
+ t.Run(model.name+"/"+tier.name, func(t *testing.T) {
+ svc := NewBillingService(&config.Config{}, nil)
+ serviceTier := ""
+ if tier.name != "standard" {
+ serviceTier = tier.name
+ }
+ cost, err := svc.CalculateCostWithServiceTier(model.name, tokens, 1, serviceTier)
+ require.NoError(t, err)
+ require.InDelta(t, float64(tokens.InputTokens)*model.input*tier.priceScale*2, cost.InputCost, 1e-12)
+ require.InDelta(t, float64(tokens.CacheCreationTokens)*model.cacheWrite*tier.priceScale*2, cost.CacheCreationCost, 1e-12)
+ require.InDelta(t, float64(tokens.CacheReadTokens)*model.cached*tier.priceScale*2, cost.CacheReadCost, 1e-12)
+ require.InDelta(t, float64(tokens.OutputTokens)*model.output*tier.priceScale*1.5, cost.OutputCost, 1e-12)
+ })
+ }
+ }
+}
+
+func TestBillingService_GPT56LongContextBoundaryIsExclusive(t *testing.T) {
+ svc := NewBillingService(&config.Config{}, nil)
+ tokens := UsageTokens{InputTokens: 100000, CacheCreationTokens: 100000, CacheReadTokens: 72000, OutputTokens: 10}
+
+ cost, err := svc.CalculateCost("gpt-5.6-sol", tokens, 1)
require.NoError(t, err)
require.InDelta(t, 100000*5e-6, cost.InputCost, 1e-12)
- require.InDelta(t, 173000*6.25e-6, cost.CacheCreationCost, 1e-12)
+ require.InDelta(t, 100000*6.25e-6, cost.CacheCreationCost, 1e-12)
+ require.InDelta(t, 72000*0.5e-6, cost.CacheReadCost, 1e-12)
require.InDelta(t, 10*30e-6, cost.OutputCost, 1e-12)
}
+func TestPricingService_BareGPT56AliasDeterministicallyUsesSol(t *testing.T) {
+ pricingSvc := &PricingService{pricingData: map[string]*LiteLLMModelPricing{
+ "gpt-5.6-sol": {InputCostPerToken: 5e-6},
+ "gpt-5.6-terra": {InputCostPerToken: 2.5e-6},
+ "gpt-5.6-luna": {InputCostPerToken: 1e-6},
+ "gpt-5.4": {InputCostPerToken: 2.5e-6},
+ }}
+
+ for i := 0; i < 100; i++ {
+ for _, alias := range []string{"gpt-5.6", "openai/gpt-5.6"} {
+ pricing := pricingSvc.GetModelPricing(alias)
+ require.NotNil(t, pricing)
+ require.InDelta(t, 5e-6, pricing.InputCostPerToken, 1e-12, "iteration=%d alias=%s", i, alias)
+ }
+ }
+
+ billingSvc := NewBillingService(&config.Config{}, pricingSvc)
+ for _, alias := range []string{"gpt-5.6", "openai/gpt-5.6"} {
+ pricing, err := billingSvc.GetModelPricing(alias)
+ require.NoError(t, err)
+ require.InDelta(t, 5e-6, pricing.InputPricePerToken, 1e-12)
+ require.InDelta(t, 6.25e-6, pricing.CacheCreationPricePerToken, 1e-12)
+ }
+}
+
func TestDefaultPricingIncludesOfficialGPT56Rates(t *testing.T) {
data, err := os.ReadFile(filepath.Join("..", "..", "resources", "model-pricing", "model_prices_and_context_window.json"))
require.NoError(t, err)
@@ -140,7 +210,9 @@ func TestDefaultPricingIncludesOfficialGPT56Rates(t *testing.T) {
require.InDelta(t, tt.cachedPriority, pricing.CacheReadPricePerTokenPriority, 1e-12)
require.InDelta(t, tt.cacheWritePriority, pricing.CacheCreationPricePerTokenPriority, 1e-12)
require.InDelta(t, tt.outputPriority, pricing.OutputPricePerTokenPriority, 1e-12)
- require.Zero(t, pricing.LongContextInputThreshold)
+ require.Equal(t, 272000, pricing.LongContextInputThreshold)
+ require.InDelta(t, 2.0, pricing.LongContextInputMultiplier, 1e-12)
+ require.InDelta(t, 1.5, pricing.LongContextOutputMultiplier, 1e-12)
})
}
}
@@ -181,7 +253,9 @@ func assertGPT56FallbackPricing(t *testing.T, pricing *ModelPricing, input, cach
require.InDelta(t, cached, pricing.CacheReadPricePerToken, 1e-12)
require.InDelta(t, cacheWrite, pricing.CacheCreationPricePerToken, 1e-12)
require.InDelta(t, output, pricing.OutputPricePerToken, 1e-12)
- require.Zero(t, pricing.LongContextInputThreshold)
+ 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 TestParsePricingData_KeepsImageOnlyPricing(t *testing.T) {
diff --git a/backend/migrations/173_allow_cyber_blocked_usage_request_type.sql b/backend/migrations/173_allow_cyber_blocked_usage_request_type.sql
new file mode 100644
index 0000000000..57b1672fa8
--- /dev/null
+++ b/backend/migrations/173_allow_cyber_blocked_usage_request_type.sql
@@ -0,0 +1,8 @@
+-- Cyber-policy blocks are recorded as request_type=4 so they remain visible in
+-- usage audits without being confused with legacy request_type=0 rows.
+ALTER TABLE usage_logs
+ DROP CONSTRAINT IF EXISTS usage_logs_request_type_check;
+
+ALTER TABLE usage_logs
+ ADD CONSTRAINT usage_logs_request_type_check
+ CHECK (request_type IN (0, 1, 2, 3, 4)) NOT VALID;
diff --git a/backend/migrations/auth_identity_payment_migrations_regression_test.go b/backend/migrations/auth_identity_payment_migrations_regression_test.go
index 14dab99160..68538cdd07 100644
--- a/backend/migrations/auth_identity_payment_migrations_regression_test.go
+++ b/backend/migrations/auth_identity_payment_migrations_regression_test.go
@@ -213,3 +213,30 @@ func TestMigration154aAddsSparkShadowIndexesConcurrently(t *testing.T) {
require.Contains(t, sql, "quota_dimension = 'spark'")
require.Contains(t, sql, "deleted_at IS NULL")
}
+
+func TestMigration173AllowsCyberBlockedUsageRequestType(t *testing.T) {
+ entries, err := FS.ReadDir(".")
+ require.NoError(t, err)
+
+ previousIndex := -1
+ currentIndex := -1
+ for i, entry := range entries {
+ switch entry.Name() {
+ case "172_video_per_second_billing_metadata.sql":
+ previousIndex = i
+ case "173_allow_cyber_blocked_usage_request_type.sql":
+ currentIndex = i
+ }
+ }
+ require.NotEqual(t, -1, previousIndex)
+ require.NotEqual(t, -1, currentIndex)
+ require.Less(t, previousIndex, currentIndex)
+
+ content, err := FS.ReadFile("173_allow_cyber_blocked_usage_request_type.sql")
+ require.NoError(t, err)
+
+ sql := string(content)
+ require.Contains(t, sql, "DROP CONSTRAINT IF EXISTS usage_logs_request_type_check")
+ require.Contains(t, sql, "ADD CONSTRAINT usage_logs_request_type_check")
+ require.Contains(t, sql, "CHECK (request_type IN (0, 1, 2, 3, 4)) NOT VALID")
+}
diff --git a/backend/resources/model-pricing/model_prices_and_context_window.json b/backend/resources/model-pricing/model_prices_and_context_window.json
index 439988a53d..a762b2145f 100644
--- a/backend/resources/model-pricing/model_prices_and_context_window.json
+++ b/backend/resources/model-pricing/model_prices_and_context_window.json
@@ -4972,6 +4972,9 @@
"input_cost_per_token_batches": 2.5e-06,
"input_cost_per_token_flex": 2.5e-06,
"input_cost_per_token_priority": 1e-05,
+ "long_context_input_token_threshold": 272000,
+ "long_context_input_cost_multiplier": 2.0,
+ "long_context_output_cost_multiplier": 1.5,
"litellm_provider": "openai",
"max_input_tokens": 1050000,
"max_output_tokens": 128000,
@@ -5021,6 +5024,9 @@
"input_cost_per_token_batches": 1.25e-06,
"input_cost_per_token_flex": 1.25e-06,
"input_cost_per_token_priority": 5e-06,
+ "long_context_input_token_threshold": 272000,
+ "long_context_input_cost_multiplier": 2.0,
+ "long_context_output_cost_multiplier": 1.5,
"litellm_provider": "openai",
"max_input_tokens": 1050000,
"max_output_tokens": 128000,
@@ -5070,6 +5076,9 @@
"input_cost_per_token_batches": 5e-07,
"input_cost_per_token_flex": 5e-07,
"input_cost_per_token_priority": 2e-06,
+ "long_context_input_token_threshold": 272000,
+ "long_context_input_cost_multiplier": 2.0,
+ "long_context_output_cost_multiplier": 1.5,
"litellm_provider": "openai",
"max_input_tokens": 1050000,
"max_output_tokens": 128000,
diff --git a/frontend/src/components/keys/UseKeyModal.vue b/frontend/src/components/keys/UseKeyModal.vue
index 5900644814..d3bca5188a 100644
--- a/frontend/src/components/keys/UseKeyModal.vue
+++ b/frontend/src/components/keys/UseKeyModal.vue
@@ -636,6 +636,23 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
xhigh: {}
}
},
+ 'gpt-5.6': {
+ name: 'GPT-5.6 (Sol)',
+ limit: {
+ context: 1050000,
+ output: 128000
+ },
+ options: {
+ store: false
+ },
+ variants: {
+ low: {},
+ medium: {},
+ high: {},
+ xhigh: {},
+ max: {}
+ }
+ },
'gpt-5.6-sol': {
name: 'GPT-5.6 Sol',
limit: {
@@ -649,7 +666,8 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
low: {},
medium: {},
high: {},
- xhigh: {}
+ xhigh: {},
+ max: {}
}
},
'gpt-5.6-terra': {
@@ -665,7 +683,8 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
low: {},
medium: {},
high: {},
- xhigh: {}
+ xhigh: {},
+ max: {}
}
},
'gpt-5.6-luna': {
@@ -681,7 +700,8 @@ function generateOpenCodeConfig(platform: string, baseUrl: string, apiKey: strin
low: {},
medium: {},
high: {},
- xhigh: {}
+ xhigh: {},
+ max: {}
}
},
'gpt-5.5': {
diff --git a/frontend/src/components/keys/__tests__/UseKeyModal.spec.ts b/frontend/src/components/keys/__tests__/UseKeyModal.spec.ts
index 8ad9731baa..f8395dc75f 100644
--- a/frontend/src/components/keys/__tests__/UseKeyModal.spec.ts
+++ b/frontend/src/components/keys/__tests__/UseKeyModal.spec.ts
@@ -123,6 +123,43 @@ describe('UseKeyModal', () => {
expect(codeBlock.text()).not.toContain('"name": "GPT-5.4 Nano"')
})
+ it('renders GPT-5.6 alias and max variants in OpenCode config', async () => {
+ const wrapper = mount(UseKeyModal, {
+ props: {
+ show: true,
+ apiKey: 'sk-test',
+ baseUrl: 'https://example.com/v1',
+ platform: 'openai'
+ },
+ global: {
+ stubs: {
+ BaseDialog: {
+ template: '
'
+ },
+ Icon: {
+ template: ''
+ }
+ }
+ }
+ })
+
+ const opencodeTab = wrapper.findAll('button').find((button) =>
+ button.text().includes('keys.useKeyModal.cliTabs.opencode')
+ )
+ expect(opencodeTab).toBeDefined()
+ await opencodeTab!.trigger('click')
+ await nextTick()
+
+ const parsed = JSON.parse(wrapper.find('pre code').text())
+ const models = parsed.provider.openai.models
+ for (const model of ['gpt-5.6', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna']) {
+ expect(models[model]).toBeDefined()
+ expect(models[model].variants).toHaveProperty('max')
+ expect(models[model].variants).toHaveProperty('xhigh')
+ }
+ expect(models['gpt-5.6'].name).toBe('GPT-5.6 (Sol)')
+ })
+
it('renders Claude Fable 5 OpenCode config with adaptive thinking', async () => {
const wrapper = mount(UseKeyModal, {
props: {
diff --git a/frontend/src/composables/__tests__/useModelWhitelist.spec.ts b/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
index a9de605c28..cd186213d6 100644
--- a/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
+++ b/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
@@ -14,6 +14,7 @@ describe('useModelWhitelist', () => {
expect(models).toContain('gpt-5.4-mini')
expect(models).toContain('gpt-5.4-2026-03-05')
expect(models).toContain('codex-auto-review')
+ expect(models).toContain('gpt-5.6')
})
it('openai 模型列表不再暴露已下线的 ChatGPT 登录 Codex 模型', () => {
diff --git a/frontend/src/composables/useModelWhitelist.ts b/frontend/src/composables/useModelWhitelist.ts
index cce3df6a52..bbd98711fd 100644
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -8,7 +8,7 @@ const openaiModels = [
'gpt-5.2', 'gpt-5.2-2025-12-11', 'gpt-5.2-chat-latest',
'gpt-5.2-pro', 'gpt-5.2-pro-2025-12-11',
// GPT-5.6 系列
- 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna',
+ 'gpt-5.6', 'gpt-5.6-sol', 'gpt-5.6-terra', 'gpt-5.6-luna',
// GPT-5.5 系列
'gpt-5.5',
// GPT-5.4 系列
@@ -280,6 +280,7 @@ const openaiPresetMappings = [
{ label: 'o3', from: 'o3', to: 'o3', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' },
{ 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-5.2', 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-5.6', from: 'gpt-5.6', to: 'gpt-5.6', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' },
{ label: 'GPT-5.6 Sol', from: 'gpt-5.6-sol', to: 'gpt-5.6-sol', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
{ label: 'GPT-5.6 Terra', from: 'gpt-5.6-terra', to: 'gpt-5.6-terra', color: 'bg-lime-100 text-lime-700 hover:bg-lime-200 dark:bg-lime-900/30 dark:text-lime-400' },
{ label: 'GPT-5.6 Luna', from: 'gpt-5.6-luna', to: 'gpt-5.6-luna', color: 'bg-sky-100 text-sky-700 hover:bg-sky-200 dark:bg-sky-900/30 dark:text-sky-400' },