Merge 6e634b85cf into 8df4a2670b

closes #2024, closes #2145

Dockerfile

@@ -44,4 +44,4 @@ COPY --from=builder2 /build/one-api /
 
 EXPOSE 3000
 WORKDIR /data
-ENTRYPOINT ["/one-api"]
+ENTRYPOINT ["/one-api"]

README.md

@@ -72,7 +72,7 @@ _✨ 通过标准的 OpenAI API 格式访问所有的大模型，开箱即用
    + [x] [Anthropic Claude 系列模型](https://anthropic.com) (支持 AWS Claude)
    + [x] [Google PaLM2/Gemini 系列模型](https://developers.generativeai.google)
    + [x] [Mistral 系列模型](https://mistral.ai/)
-   + [x] [字节跳动豆包大模型](https://console.volcengine.com/ark/region:ark+cn-beijing/model)
+   + [x] [字节跳动豆包大模型（火山引擎）](https://www.volcengine.com/experience/ark?utm_term=202502dsinvite&ac=DSASUQY5&rc=2QXCA1VI)
    + [x] [百度文心一言系列模型](https://cloud.baidu.com/doc/WENXINWORKSHOP/index.html)
    + [x] [阿里通义千问系列模型](https://help.aliyun.com/document_detail/2400395.html)
    + [x] [讯飞星火认知大模型](https://www.xfyun.cn/doc/spark/Web.html)
@@ -385,7 +385,7 @@ graph LR
    + 例子：`NODE_TYPE=slave`
 9. `CHANNEL_UPDATE_FREQUENCY`：设置之后将定期更新渠道余额，单位为分钟，未设置则不进行更新。
    + 例子：`CHANNEL_UPDATE_FREQUENCY=1440`
-10. `CHANNEL_TEST_FREQUENCY`：设置之后将定期检查渠道，单位为分钟，未设置则不进行检查。 
+10. `CHANNEL_TEST_FREQUENCY`：设置之后将定期检查渠道，单位为分钟，未设置则不进行检查。
    +例子：`CHANNEL_TEST_FREQUENCY=1440`
 11. `POLLING_INTERVAL`：批量更新渠道余额以及测试可用性时的请求间隔，单位为秒，默认无间隔。
     + 例子：`POLLING_INTERVAL=5`

common/config/config.go

@@ -164,3 +164,6 @@ var UserContentRequestTimeout = env.Int("USER_CONTENT_REQUEST_TIMEOUT", 30)
 
 var EnforceIncludeUsage = env.Bool("ENFORCE_INCLUDE_USAGE", false)
 var TestPrompt = env.String("TEST_PROMPT", "Output only your specific model name with no additional text.")
+
+// OpenrouterProviderSort is used to determine the order of the providers in the openrouter
+var OpenrouterProviderSort = env.String("OPENROUTER_PROVIDER_SORT", "")

common/conv/any.go

@@ -1,6 +1,9 @@
 package conv
 
 func AsString(v any) string {
-	str, _ := v.(string)
-	return str
+	if str, ok := v.(string); ok {
+		return str
+	}
+
+	return ""
 }

relay/adaptor/ali/constants.go

@@ -14,10 +14,14 @@ var ModelList = []string{
 	"qwen2-72b-instruct", "qwen2-57b-a14b-instruct", "qwen2-7b-instruct", "qwen2-1.5b-instruct", "qwen2-0.5b-instruct",
 	"qwen1.5-110b-chat", "qwen1.5-72b-chat", "qwen1.5-32b-chat", "qwen1.5-14b-chat", "qwen1.5-7b-chat", "qwen1.5-1.8b-chat", "qwen1.5-0.5b-chat",
 	"qwen-72b-chat", "qwen-14b-chat", "qwen-7b-chat", "qwen-1.8b-chat", "qwen-1.8b-longcontext-chat",
+	"qvq-72b-preview",
+	"qwen2.5-vl-72b-instruct", "qwen2.5-vl-7b-instruct", "qwen2.5-vl-2b-instruct", "qwen2.5-vl-1b-instruct", "qwen2.5-vl-0.5b-instruct",
 	"qwen2-vl-7b-instruct", "qwen2-vl-2b-instruct", "qwen-vl-v1", "qwen-vl-chat-v1",
 	"qwen2-audio-instruct", "qwen-audio-chat",
 	"qwen2.5-math-72b-instruct", "qwen2.5-math-7b-instruct", "qwen2.5-math-1.5b-instruct", "qwen2-math-72b-instruct", "qwen2-math-7b-instruct", "qwen2-math-1.5b-instruct",
 	"qwen2.5-coder-32b-instruct", "qwen2.5-coder-14b-instruct", "qwen2.5-coder-7b-instruct", "qwen2.5-coder-3b-instruct", "qwen2.5-coder-1.5b-instruct", "qwen2.5-coder-0.5b-instruct",
 	"text-embedding-v1", "text-embedding-v3", "text-embedding-v2", "text-embedding-async-v2", "text-embedding-async-v1",
 	"ali-stable-diffusion-xl", "ali-stable-diffusion-v1.5", "wanx-v1",
+	"qwen-mt-plus", "qwen-mt-turbo",
+	"deepseek-r1", "deepseek-v3", "deepseek-r1-distill-qwen-1.5b", "deepseek-r1-distill-qwen-7b", "deepseek-r1-distill-qwen-14b", "deepseek-r1-distill-qwen-32b", "deepseek-r1-distill-llama-8b", "deepseek-r1-distill-llama-70b",
 }

relay/adaptor/anthropic/adaptor.go

@@ -36,8 +36,8 @@ func (a *Adaptor) SetupRequestHeader(c *gin.Context, req *http.Request, meta *me
 
 	// https://x.com/alexalbert__/status/1812921642143900036
 	// claude-3-5-sonnet can support 8k context
-	if strings.HasPrefix(meta.ActualModelName, "claude-3-5-sonnet") {
-		req.Header.Set("anthropic-beta", "max-tokens-3-5-sonnet-2024-07-15")
+	if strings.HasPrefix(meta.ActualModelName, "claude-3-7-sonnet") {
+		req.Header.Set("anthropic-beta", "output-128k-2025-02-19")
 	}
 
 	return nil
@@ -47,7 +47,7 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
-	return ConvertRequest(*request), nil
+	return ConvertRequest(c, *request)
 }
 
 func (a *Adaptor) ConvertImageRequest(request *model.ImageRequest) (any, error) {

relay/adaptor/anthropic/constants.go

@@ -3,11 +3,13 @@ package anthropic
 var ModelList = []string{
 	"claude-instant-1.2", "claude-2.0", "claude-2.1",
 	"claude-3-haiku-20240307",
-	"claude-3-5-haiku-20241022",
 	"claude-3-5-haiku-latest",
+	"claude-3-5-haiku-20241022",
 	"claude-3-sonnet-20240229",
 	"claude-3-opus-20240229",
+	"claude-3-5-sonnet-latest",
 	"claude-3-5-sonnet-20240620",
 	"claude-3-5-sonnet-20241022",
-	"claude-3-5-sonnet-latest",
+	"claude-3-7-sonnet-latest",
+	"claude-3-7-sonnet-20250219",
 }

relay/adaptor/anthropic/main.go

@@ -2,18 +2,21 @@ package anthropic
 
 import (
 	"bufio"
+	"context"
 	"encoding/json"
 	"fmt"
-	"github.com/songquanpeng/one-api/common/render"
 	"io"
+	"math"
 	"net/http"
 	"strings"
 
 	"github.com/gin-gonic/gin"
+	"github.com/pkg/errors"
 	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/common/helper"
 	"github.com/songquanpeng/one-api/common/image"
 	"github.com/songquanpeng/one-api/common/logger"
+	"github.com/songquanpeng/one-api/common/render"
 	"github.com/songquanpeng/one-api/relay/adaptor/openai"
 	"github.com/songquanpeng/one-api/relay/model"
 )
@@ -36,7 +39,16 @@ func stopReasonClaude2OpenAI(reason *string) string {
 	}
 }
 
-func ConvertRequest(textRequest model.GeneralOpenAIRequest) *Request {
+// isModelSupportThinking is used to check if the model supports extended thinking
+func isModelSupportThinking(model string) bool {
+	if strings.Contains(model, "claude-3-7-sonnet") {
+		return true
+	}
+
+	return false
+}
+
+func ConvertRequest(c *gin.Context, textRequest model.GeneralOpenAIRequest) (*Request, error) {
 	claudeTools := make([]Tool, 0, len(textRequest.Tools))
 
 	for _, tool := range textRequest.Tools {
@@ -61,7 +73,27 @@ func ConvertRequest(textRequest model.GeneralOpenAIRequest) *Request {
 		TopK:        textRequest.TopK,
 		Stream:      textRequest.Stream,
 		Tools:       claudeTools,
+		Thinking:    textRequest.Thinking,
 	}
+
+	if isModelSupportThinking(textRequest.Model) &&
+		c.Request.URL.Query().Has("thinking") && claudeRequest.Thinking == nil {
+		claudeRequest.Thinking = &model.Thinking{
+			Type:         "enabled",
+			BudgetTokens: int(math.Min(1024, float64(claudeRequest.MaxTokens/2))),
+		}
+	}
+
+	if isModelSupportThinking(textRequest.Model) &&
+		claudeRequest.Thinking != nil {
+		if claudeRequest.MaxTokens <= 1024 {
+			return nil, errors.New("max_tokens must be greater than 1024 when using extended thinking")
+		}
+
+		// top_p must be nil when using extended thinking
+		claudeRequest.TopP = nil
+	}
+
 	if len(claudeTools) > 0 {
 		claudeToolChoice := struct {
 			Type string `json:"type"`
@@ -142,13 +174,14 @@ func ConvertRequest(textRequest model.GeneralOpenAIRequest) *Request {
 		claudeMessage.Content = contents
 		claudeRequest.Messages = append(claudeRequest.Messages, claudeMessage)
 	}
-	return &claudeRequest
+	return &claudeRequest, nil
 }
 
 // https://docs.anthropic.com/claude/reference/messages-streaming
 func StreamResponseClaude2OpenAI(claudeResponse *StreamResponse) (*openai.ChatCompletionsStreamResponse, *Response) {
 	var response *Response
 	var responseText string
+	var reasoningText string
 	var stopReason string
 	tools := make([]model.Tool, 0)
 
@@ -158,6 +191,10 @@ func StreamResponseClaude2OpenAI(claudeResponse *StreamResponse) (*openai.ChatCo
 	case "content_block_start":
 		if claudeResponse.ContentBlock != nil {
 			responseText = claudeResponse.ContentBlock.Text
+			if claudeResponse.ContentBlock.Thinking != nil {
+				reasoningText = *claudeResponse.ContentBlock.Thinking
+			}
+
 			if claudeResponse.ContentBlock.Type == "tool_use" {
 				tools = append(tools, model.Tool{
 					Id:   claudeResponse.ContentBlock.Id,
@@ -172,6 +209,10 @@ func StreamResponseClaude2OpenAI(claudeResponse *StreamResponse) (*openai.ChatCo
 	case "content_block_delta":
 		if claudeResponse.Delta != nil {
 			responseText = claudeResponse.Delta.Text
+			if claudeResponse.Delta.Thinking != nil {
+				reasoningText = *claudeResponse.Delta.Thinking
+			}
+
 			if claudeResponse.Delta.Type == "input_json_delta" {
 				tools = append(tools, model.Tool{
 					Function: model.Function{
@@ -189,9 +230,20 @@ func StreamResponseClaude2OpenAI(claudeResponse *StreamResponse) (*openai.ChatCo
 		if claudeResponse.Delta != nil && claudeResponse.Delta.StopReason != nil {
 			stopReason = *claudeResponse.Delta.StopReason
 		}
+	case "thinking_delta":
+		if claudeResponse.Delta != nil && claudeResponse.Delta.Thinking != nil {
+			reasoningText = *claudeResponse.Delta.Thinking
+		}
+	case "ping",
+		"message_stop",
+		"content_block_stop":
+	default:
+		logger.SysErrorf("unknown stream response type %q", claudeResponse.Type)
 	}
+
 	var choice openai.ChatCompletionsStreamResponseChoice
 	choice.Delta.Content = responseText
+	choice.Delta.Reasoning = &reasoningText
 	if len(tools) > 0 {
 		choice.Delta.Content = nil // compatible with other OpenAI derivative applications, like LobeOpenAICompatibleFactory ...
 		choice.Delta.ToolCalls = tools
@@ -209,11 +261,23 @@ func StreamResponseClaude2OpenAI(claudeResponse *StreamResponse) (*openai.ChatCo
 
 func ResponseClaude2OpenAI(claudeResponse *Response) *openai.TextResponse {
 	var responseText string
-	if len(claudeResponse.Content) > 0 {
-		responseText = claudeResponse.Content[0].Text
-	}
+	var reasoningText string
+
 	tools := make([]model.Tool, 0)
 	for _, v := range claudeResponse.Content {
+		switch v.Type {
+		case "thinking":
+			if v.Thinking != nil {
+				reasoningText += *v.Thinking
+			} else {
+				logger.Errorf(context.Background(), "thinking is nil in response")
+			}
+		case "text":
+			responseText += v.Text
+		default:
+			logger.Warnf(context.Background(), "unknown response type %q", v.Type)
+		}
+
 		if v.Type == "tool_use" {
 			args, _ := json.Marshal(v.Input)
 			tools = append(tools, model.Tool{
@@ -226,11 +290,13 @@ func ResponseClaude2OpenAI(claudeResponse *Response) *openai.TextResponse {
 			})
 		}
 	}
+
 	choice := openai.TextResponseChoice{
 		Index: 0,
 		Message: model.Message{
 			Role:      "assistant",
 			Content:   responseText,
+			Reasoning: &reasoningText,
 			Name:      nil,
 			ToolCalls: tools,
 		},
@@ -277,6 +343,8 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC
 		data = strings.TrimPrefix(data, "data:")
 		data = strings.TrimSpace(data)
 
+		logger.Debugf(c.Request.Context(), "stream <- %q\n", data)
+
 		var claudeResponse StreamResponse
 		err := json.Unmarshal([]byte(data), &claudeResponse)
 		if err != nil {
@@ -344,6 +412,9 @@ func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName st
 	if err != nil {
 		return openai.ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 	}
+
+	logger.Debugf(c.Request.Context(), "response <- %s\n", string(responseBody))
+
 	var claudeResponse Response
 	err = json.Unmarshal(responseBody, &claudeResponse)
 	if err != nil {

relay/adaptor/anthropic/model.go

@@ -1,5 +1,7 @@
 package anthropic
 
+import "github.com/songquanpeng/one-api/relay/model"
+
 // https://docs.anthropic.com/claude/reference/messages_post
 
 type Metadata struct {
@@ -22,6 +24,9 @@ type Content struct {
 	Input     any    `json:"input,omitempty"`
 	Content   string `json:"content,omitempty"`
 	ToolUseId string `json:"tool_use_id,omitempty"`
+	// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#implementing-extended-thinking
+	Thinking  *string `json:"thinking,omitempty"`
+	Signature *string `json:"signature,omitempty"`
 }
 
 type Message struct {
@@ -54,6 +59,7 @@ type Request struct {
 	Tools         []Tool    `json:"tools,omitempty"`
 	ToolChoice    any       `json:"tool_choice,omitempty"`
 	//Metadata    `json:"metadata,omitempty"`
+	Thinking *model.Thinking `json:"thinking,omitempty"`
 }
 
 type Usage struct {
@@ -84,6 +90,8 @@ type Delta struct {
 	PartialJson  string  `json:"partial_json,omitempty"`
 	StopReason   *string `json:"stop_reason"`
 	StopSequence *string `json:"stop_sequence"`
+	Thinking     *string `json:"thinking,omitempty"`
+	Signature    *string `json:"signature,omitempty"`
 }
 
 type StreamResponse struct {

relay/adaptor/aws/claude/adapter.go

@@ -21,7 +21,11 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 		return nil, errors.New("request is nil")
 	}
 
-	claudeReq := anthropic.ConvertRequest(*request)
+	claudeReq, err := anthropic.ConvertRequest(c, *request)
+	if err != nil {
+		return nil, errors.Wrap(err, "convert request")
+	}
+
 	c.Set(ctxkey.RequestModel, request.Model)
 	c.Set(ctxkey.ConvertedRequest, claudeReq)
 	return claudeReq, nil

relay/adaptor/aws/claude/main.go

@@ -36,6 +36,8 @@ var AwsModelIDMap = map[string]string{
 	"claude-3-5-sonnet-20241022": "anthropic.claude-3-5-sonnet-20241022-v2:0",
 	"claude-3-5-sonnet-latest":   "anthropic.claude-3-5-sonnet-20241022-v2:0",
 	"claude-3-5-haiku-20241022":  "anthropic.claude-3-5-haiku-20241022-v1:0",
+	"claude-3-7-sonnet-latest":   "anthropic.claude-3-7-sonnet-20250219-v1:0",
+	"claude-3-7-sonnet-20250219": "anthropic.claude-3-7-sonnet-20250219-v1:0",
 }
 
 func awsModelID(requestModel string) (string, error) {
@@ -47,13 +49,14 @@ func awsModelID(requestModel string) (string, error) {
 }
 
 func Handler(c *gin.Context, awsCli *bedrockruntime.Client, modelName string) (*relaymodel.ErrorWithStatusCode, *relaymodel.Usage) {
-	awsModelId, err := awsModelID(c.GetString(ctxkey.RequestModel))
+	awsModelID, err := awsModelID(c.GetString(ctxkey.RequestModel))
 	if err != nil {
 		return utils.WrapErr(errors.Wrap(err, "awsModelID")), nil
 	}
 
+	awsModelID = utils.ConvertModelID2CrossRegionProfile(awsModelID, awsCli.Options().Region)
 	awsReq := &bedrockruntime.InvokeModelInput{
-		ModelId:     aws.String(awsModelId),
+		ModelId:     aws.String(awsModelID),
 		Accept:      aws.String("application/json"),
 		ContentType: aws.String("application/json"),
 	}
@@ -101,13 +104,14 @@ func Handler(c *gin.Context, awsCli *bedrockruntime.Client, modelName string) (*
 
 func StreamHandler(c *gin.Context, awsCli *bedrockruntime.Client) (*relaymodel.ErrorWithStatusCode, *relaymodel.Usage) {
 	createdTime := helper.GetTimestamp()
-	awsModelId, err := awsModelID(c.GetString(ctxkey.RequestModel))
+	awsModelID, err := awsModelID(c.GetString(ctxkey.RequestModel))
 	if err != nil {
 		return utils.WrapErr(errors.Wrap(err, "awsModelID")), nil
 	}
 
+	awsModelID = utils.ConvertModelID2CrossRegionProfile(awsModelID, awsCli.Options().Region)
 	awsReq := &bedrockruntime.InvokeModelWithResponseStreamInput{
-		ModelId:     aws.String(awsModelId),
+		ModelId:     aws.String(awsModelID),
 		Accept:      aws.String("application/json"),
 		ContentType: aws.String("application/json"),
 	}

relay/adaptor/aws/claude/model.go

@@ -1,6 +1,9 @@
 package aws
 
-import "github.com/songquanpeng/one-api/relay/adaptor/anthropic"
+import (
+	"github.com/songquanpeng/one-api/relay/adaptor/anthropic"
+	"github.com/songquanpeng/one-api/relay/model"
+)
 
 // Request is the request to AWS Claude
 //
@@ -17,4 +20,5 @@ type Request struct {
 	StopSequences    []string            `json:"stop_sequences,omitempty"`
 	Tools            []anthropic.Tool    `json:"tools,omitempty"`
 	ToolChoice       any                 `json:"tool_choice,omitempty"`
+	Thinking         *model.Thinking     `json:"thinking,omitempty"`
 }

relay/adaptor/aws/llama3/main.go

@@ -70,13 +70,14 @@ func ConvertRequest(textRequest relaymodel.GeneralOpenAIRequest) *Request {
 }
 
 func Handler(c *gin.Context, awsCli *bedrockruntime.Client, modelName string) (*relaymodel.ErrorWithStatusCode, *relaymodel.Usage) {
-	awsModelId, err := awsModelID(c.GetString(ctxkey.RequestModel))
+	awsModelID, err := awsModelID(c.GetString(ctxkey.RequestModel))
 	if err != nil {
 		return utils.WrapErr(errors.Wrap(err, "awsModelID")), nil
 	}
 
+	awsModelID = utils.ConvertModelID2CrossRegionProfile(awsModelID, awsCli.Options().Region)
 	awsReq := &bedrockruntime.InvokeModelInput{
-		ModelId:     aws.String(awsModelId),
+		ModelId:     aws.String(awsModelID),
 		Accept:      aws.String("application/json"),
 		ContentType: aws.String("application/json"),
 	}
@@ -140,13 +141,14 @@ func ResponseLlama2OpenAI(llamaResponse *Response) *openai.TextResponse {
 
 func StreamHandler(c *gin.Context, awsCli *bedrockruntime.Client) (*relaymodel.ErrorWithStatusCode, *relaymodel.Usage) {
 	createdTime := helper.GetTimestamp()
-	awsModelId, err := awsModelID(c.GetString(ctxkey.RequestModel))
+	awsModelID, err := awsModelID(c.GetString(ctxkey.RequestModel))
 	if err != nil {
 		return utils.WrapErr(errors.Wrap(err, "awsModelID")), nil
 	}
 
+	awsModelID = utils.ConvertModelID2CrossRegionProfile(awsModelID, awsCli.Options().Region)
 	awsReq := &bedrockruntime.InvokeModelWithResponseStreamInput{
-		ModelId:     aws.String(awsModelId),
+		ModelId:     aws.String(awsModelID),
 		Accept:      aws.String("application/json"),
 		ContentType: aws.String("application/json"),
 	}

relay/adaptor/aws/utils/consts.go

@@ -0,0 +1,75 @@
+package utils
+
+import (
+	"context"
+	"slices"
+	"strings"
+
+	"github.com/songquanpeng/one-api/common/logger"
+)
+
+// CrossRegionInferences is a list of model IDs that support cross-region inference.
+//
+// https://docs.aws.amazon.com/bedrock/latest/userguide/inference-profiles-support.html
+//
+// document.querySelectorAll('pre.programlisting code').forEach((e) => {console.log(e.innerHTML)})
+var CrossRegionInferences = []string{
+	"us.amazon.nova-lite-v1:0",
+	"us.amazon.nova-micro-v1:0",
+	"us.amazon.nova-pro-v1:0",
+	"us.anthropic.claude-3-5-haiku-20241022-v1:0",
+	"us.anthropic.claude-3-5-sonnet-20240620-v1:0",
+	"us.anthropic.claude-3-5-sonnet-20241022-v2:0",
+	"us.anthropic.claude-3-7-sonnet-20250219-v1:0",
+	"us.anthropic.claude-3-haiku-20240307-v1:0",
+	"us.anthropic.claude-3-opus-20240229-v1:0",
+	"us.anthropic.claude-3-sonnet-20240229-v1:0",
+	"us.meta.llama3-1-405b-instruct-v1:0",
+	"us.meta.llama3-1-70b-instruct-v1:0",
+	"us.meta.llama3-1-8b-instruct-v1:0",
+	"us.meta.llama3-2-11b-instruct-v1:0",
+	"us.meta.llama3-2-1b-instruct-v1:0",
+	"us.meta.llama3-2-3b-instruct-v1:0",
+	"us.meta.llama3-2-90b-instruct-v1:0",
+	"us.meta.llama3-3-70b-instruct-v1:0",
+	"us-gov.anthropic.claude-3-5-sonnet-20240620-v1:0",
+	"us-gov.anthropic.claude-3-haiku-20240307-v1:0",
+	"eu.amazon.nova-lite-v1:0",
+	"eu.amazon.nova-micro-v1:0",
+	"eu.amazon.nova-pro-v1:0",
+	"eu.anthropic.claude-3-5-sonnet-20240620-v1:0",
+	"eu.anthropic.claude-3-haiku-20240307-v1:0",
+	"eu.anthropic.claude-3-sonnet-20240229-v1:0",
+	"eu.meta.llama3-2-1b-instruct-v1:0",
+	"eu.meta.llama3-2-3b-instruct-v1:0",
+	"apac.amazon.nova-lite-v1:0",
+	"apac.amazon.nova-micro-v1:0",
+	"apac.amazon.nova-pro-v1:0",
+	"apac.anthropic.claude-3-5-sonnet-20240620-v1:0",
+	"apac.anthropic.claude-3-5-sonnet-20241022-v2:0",
+	"apac.anthropic.claude-3-haiku-20240307-v1:0",
+	"apac.anthropic.claude-3-sonnet-20240229-v1:0",
+}
+
+// ConvertModelID2CrossRegionProfile converts the model ID to a cross-region profile ID.
+func ConvertModelID2CrossRegionProfile(model, region string) string {
+	var regionPrefix string
+	switch prefix := strings.Split(region, "-")[0]; prefix {
+	case "us", "eu":
+		regionPrefix = prefix
+	case "ap":
+		regionPrefix = "apac"
+	default:
+		// not supported, return original model
+		return model
+	}
+
+	newModelID := regionPrefix + "." + model
+	if slices.Contains(CrossRegionInferences, newModelID) {
+		logger.Debugf(context.TODO(), "convert model %s to cross-region profile %s", model, newModelID)
+		return newModelID
+	}
+
+	// not found, return original model
+	return model
+}

relay/adaptor/openai/adaptor.go

@@ -9,6 +9,8 @@ import (
 
 	"github.com/gin-gonic/gin"
 
+	"github.com/songquanpeng/one-api/common/config"
+	"github.com/songquanpeng/one-api/common/logger"
 	"github.com/songquanpeng/one-api/relay/adaptor"
 	"github.com/songquanpeng/one-api/relay/adaptor/alibailian"
 	"github.com/songquanpeng/one-api/relay/adaptor/baiduv2"
@@ -16,6 +18,7 @@ import (
 	"github.com/songquanpeng/one-api/relay/adaptor/geminiv2"
 	"github.com/songquanpeng/one-api/relay/adaptor/minimax"
 	"github.com/songquanpeng/one-api/relay/adaptor/novita"
+	"github.com/songquanpeng/one-api/relay/adaptor/openrouter"
 	"github.com/songquanpeng/one-api/relay/channeltype"
 	"github.com/songquanpeng/one-api/relay/meta"
 	"github.com/songquanpeng/one-api/relay/model"
@@ -85,7 +88,29 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
-	if request.Stream {
+
+	meta := meta.GetByContext(c)
+	switch meta.ChannelType {
+	case channeltype.OpenRouter:
+		includeReasoning := true
+		request.IncludeReasoning = &includeReasoning
+		if request.Provider == nil || request.Provider.Sort == "" &&
+			config.OpenrouterProviderSort != "" {
+			if request.Provider == nil {
+				request.Provider = &openrouter.RequestProvider{}
+			}
+
+			request.Provider.Sort = config.OpenrouterProviderSort
+		}
+	default:
+	}
+
+	if request.Stream && !config.EnforceIncludeUsage {
+		logger.Warn(c.Request.Context(),
+			"please set ENFORCE_INCLUDE_USAGE=true to ensure accurate billing in stream mode")
+	}
+
+	if config.EnforceIncludeUsage && request.Stream {
 		// always return usage in stream mode
 		if request.StreamOptions == nil {
 			request.StreamOptions = &model.StreamOptions{}

relay/adaptor/openai/main.go

@@ -8,12 +8,11 @@ import (
 	"net/http"
 	"strings"
 
-	"github.com/songquanpeng/one-api/common/render"
-
 	"github.com/gin-gonic/gin"
 	"github.com/songquanpeng/one-api/common"
 	"github.com/songquanpeng/one-api/common/conv"
 	"github.com/songquanpeng/one-api/common/logger"
+	"github.com/songquanpeng/one-api/common/render"
 	"github.com/songquanpeng/one-api/relay/model"
 	"github.com/songquanpeng/one-api/relay/relaymode"
 )
@@ -26,6 +25,7 @@ const (
 
 func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.ErrorWithStatusCode, string, *model.Usage) {
 	responseText := ""
+	reasoningText := ""
 	scanner := bufio.NewScanner(resp.Body)
 	scanner.Split(bufio.ScanLines)
 	var usage *model.Usage
@@ -61,6 +61,13 @@ func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.E
 			}
 			render.StringData(c, data)
 			for _, choice := range streamResponse.Choices {
+				if choice.Delta.Reasoning != nil {
+					reasoningText += *choice.Delta.Reasoning
+				}
+				if choice.Delta.ReasoningContent != nil {
+					reasoningText += *choice.Delta.ReasoningContent
+				}
+
 				responseText += conv.AsString(choice.Delta.Content)
 			}
 			if streamResponse.Usage != nil {
@@ -93,7 +100,7 @@ func StreamHandler(c *gin.Context, resp *http.Response, relayMode int) (*model.E
 		return ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), "", nil
 	}
 
-	return nil, responseText, usage
+	return nil, reasoningText + responseText, usage
 }
 
 func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName string) (*model.ErrorWithStatusCode, *model.Usage) {
@@ -136,10 +143,17 @@ func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName st
 		return ErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
 	}
 
-	if textResponse.Usage.TotalTokens == 0 || (textResponse.Usage.PromptTokens == 0 && textResponse.Usage.CompletionTokens == 0) {
+	if textResponse.Usage.TotalTokens == 0 ||
+		(textResponse.Usage.PromptTokens == 0 && textResponse.Usage.CompletionTokens == 0) {
 		completionTokens := 0
 		for _, choice := range textResponse.Choices {
 			completionTokens += CountTokenText(choice.Message.StringContent(), modelName)
+			if choice.Message.Reasoning != nil {
+				completionTokens += CountToken(*choice.Message.Reasoning)
+			}
+			if choice.ReasoningContent != nil {
+				completionTokens += CountToken(*choice.ReasoningContent)
+			}
 		}
 		textResponse.Usage = model.Usage{
 			PromptTokens:     promptTokens,
@@ -147,5 +161,6 @@ func Handler(c *gin.Context, resp *http.Response, promptTokens int, modelName st
 			TotalTokens:      promptTokens + completionTokens,
 		}
 	}
+
 	return nil, &textResponse.Usage
 }

relay/adaptor/openrouter/constants.go

@@ -1,20 +1,235 @@
 package openrouter
 
 var ModelList = []string{
-	"openai/gpt-3.5-turbo",
-	"openai/chatgpt-4o-latest",
-	"openai/o1",
-	"openai/o1-preview",
-	"openai/o1-mini",
-	"openai/o3-mini",
-	"google/gemini-2.0-flash-001",
-	"google/gemini-2.0-flash-thinking-exp:free",
-	"google/gemini-2.0-flash-lite-preview-02-05:free",
-	"google/gemini-2.0-pro-exp-02-05:free",
-	"google/gemini-flash-1.5-8b",
-	"anthropic/claude-3.5-sonnet",
+	"01-ai/yi-large",
+	"aetherwiing/mn-starcannon-12b",
+	"ai21/jamba-1-5-large",
+	"ai21/jamba-1-5-mini",
+	"ai21/jamba-instruct",
+	"aion-labs/aion-1.0",
+	"aion-labs/aion-1.0-mini",
+	"aion-labs/aion-rp-llama-3.1-8b",
+	"allenai/llama-3.1-tulu-3-405b",
+	"alpindale/goliath-120b",
+	"alpindale/magnum-72b",
+	"amazon/nova-lite-v1",
+	"amazon/nova-micro-v1",
+	"amazon/nova-pro-v1",
+	"anthracite-org/magnum-v2-72b",
+	"anthracite-org/magnum-v4-72b",
+	"anthropic/claude-2",
+	"anthropic/claude-2.0",
+	"anthropic/claude-2.0:beta",
+	"anthropic/claude-2.1",
+	"anthropic/claude-2.1:beta",
+	"anthropic/claude-2:beta",
+	"anthropic/claude-3-haiku",
+	"anthropic/claude-3-haiku:beta",
+	"anthropic/claude-3-opus",
+	"anthropic/claude-3-opus:beta",
+	"anthropic/claude-3-sonnet",
+	"anthropic/claude-3-sonnet:beta",
 	"anthropic/claude-3.5-haiku",
-	"deepseek/deepseek-r1:free",
+	"anthropic/claude-3.5-haiku-20241022",
+	"anthropic/claude-3.5-haiku-20241022:beta",
+	"anthropic/claude-3.5-haiku:beta",
+	"anthropic/claude-3.5-sonnet",
+	"anthropic/claude-3.5-sonnet-20240620",
+	"anthropic/claude-3.5-sonnet-20240620:beta",
+	"anthropic/claude-3.5-sonnet:beta",
+	"cognitivecomputations/dolphin-mixtral-8x22b",
+	"cognitivecomputations/dolphin-mixtral-8x7b",
+	"cohere/command",
+	"cohere/command-r",
+	"cohere/command-r-03-2024",
+	"cohere/command-r-08-2024",
+	"cohere/command-r-plus",
+	"cohere/command-r-plus-04-2024",
+	"cohere/command-r-plus-08-2024",
+	"cohere/command-r7b-12-2024",
+	"databricks/dbrx-instruct",
+	"deepseek/deepseek-chat",
+	"deepseek/deepseek-chat-v2.5",
+	"deepseek/deepseek-chat:free",
 	"deepseek/deepseek-r1",
+	"deepseek/deepseek-r1-distill-llama-70b",
+	"deepseek/deepseek-r1-distill-llama-70b:free",
+	"deepseek/deepseek-r1-distill-llama-8b",
+	"deepseek/deepseek-r1-distill-qwen-1.5b",
+	"deepseek/deepseek-r1-distill-qwen-14b",
+	"deepseek/deepseek-r1-distill-qwen-32b",
+	"deepseek/deepseek-r1:free",
+	"eva-unit-01/eva-llama-3.33-70b",
+	"eva-unit-01/eva-qwen-2.5-32b",
+	"eva-unit-01/eva-qwen-2.5-72b",
+	"google/gemini-2.0-flash-001",
+	"google/gemini-2.0-flash-exp:free",
+	"google/gemini-2.0-flash-lite-preview-02-05:free",
+	"google/gemini-2.0-flash-thinking-exp-1219:free",
+	"google/gemini-2.0-flash-thinking-exp:free",
+	"google/gemini-2.0-pro-exp-02-05:free",
+	"google/gemini-exp-1206:free",
+	"google/gemini-flash-1.5",
+	"google/gemini-flash-1.5-8b",
+	"google/gemini-flash-1.5-8b-exp",
+	"google/gemini-pro",
+	"google/gemini-pro-1.5",
+	"google/gemini-pro-vision",
+	"google/gemma-2-27b-it",
+	"google/gemma-2-9b-it",
+	"google/gemma-2-9b-it:free",
+	"google/gemma-7b-it",
+	"google/learnlm-1.5-pro-experimental:free",
+	"google/palm-2-chat-bison",
+	"google/palm-2-chat-bison-32k",
+	"google/palm-2-codechat-bison",
+	"google/palm-2-codechat-bison-32k",
+	"gryphe/mythomax-l2-13b",
+	"gryphe/mythomax-l2-13b:free",
+	"huggingfaceh4/zephyr-7b-beta:free",
+	"infermatic/mn-inferor-12b",
+	"inflection/inflection-3-pi",
+	"inflection/inflection-3-productivity",
+	"jondurbin/airoboros-l2-70b",
+	"liquid/lfm-3b",
+	"liquid/lfm-40b",
+	"liquid/lfm-7b",
+	"mancer/weaver",
+	"meta-llama/llama-2-13b-chat",
+	"meta-llama/llama-2-70b-chat",
+	"meta-llama/llama-3-70b-instruct",
+	"meta-llama/llama-3-8b-instruct",
+	"meta-llama/llama-3-8b-instruct:free",
+	"meta-llama/llama-3.1-405b",
+	"meta-llama/llama-3.1-405b-instruct",
+	"meta-llama/llama-3.1-70b-instruct",
+	"meta-llama/llama-3.1-8b-instruct",
+	"meta-llama/llama-3.2-11b-vision-instruct",
+	"meta-llama/llama-3.2-11b-vision-instruct:free",
+	"meta-llama/llama-3.2-1b-instruct",
+	"meta-llama/llama-3.2-3b-instruct",
+	"meta-llama/llama-3.2-90b-vision-instruct",
+	"meta-llama/llama-3.3-70b-instruct",
+	"meta-llama/llama-3.3-70b-instruct:free",
+	"meta-llama/llama-guard-2-8b",
+	"microsoft/phi-3-medium-128k-instruct",
+	"microsoft/phi-3-medium-128k-instruct:free",
+	"microsoft/phi-3-mini-128k-instruct",
+	"microsoft/phi-3-mini-128k-instruct:free",
+	"microsoft/phi-3.5-mini-128k-instruct",
+	"microsoft/phi-4",
+	"microsoft/wizardlm-2-7b",
+	"microsoft/wizardlm-2-8x22b",
+	"minimax/minimax-01",
+	"mistralai/codestral-2501",
+	"mistralai/codestral-mamba",
+	"mistralai/ministral-3b",
+	"mistralai/ministral-8b",
+	"mistralai/mistral-7b-instruct",
+	"mistralai/mistral-7b-instruct-v0.1",
+	"mistralai/mistral-7b-instruct-v0.3",
+	"mistralai/mistral-7b-instruct:free",
+	"mistralai/mistral-large",
+	"mistralai/mistral-large-2407",
+	"mistralai/mistral-large-2411",
+	"mistralai/mistral-medium",
+	"mistralai/mistral-nemo",
+	"mistralai/mistral-nemo:free",
+	"mistralai/mistral-small",
+	"mistralai/mistral-small-24b-instruct-2501",
+	"mistralai/mistral-small-24b-instruct-2501:free",
+	"mistralai/mistral-tiny",
+	"mistralai/mixtral-8x22b-instruct",
+	"mistralai/mixtral-8x7b",
+	"mistralai/mixtral-8x7b-instruct",
+	"mistralai/pixtral-12b",
+	"mistralai/pixtral-large-2411",
+	"neversleep/llama-3-lumimaid-70b",
+	"neversleep/llama-3-lumimaid-8b",
+	"neversleep/llama-3-lumimaid-8b:extended",
+	"neversleep/llama-3.1-lumimaid-70b",
+	"neversleep/llama-3.1-lumimaid-8b",
+	"neversleep/noromaid-20b",
+	"nothingiisreal/mn-celeste-12b",
+	"nousresearch/hermes-2-pro-llama-3-8b",
+	"nousresearch/hermes-3-llama-3.1-405b",
+	"nousresearch/hermes-3-llama-3.1-70b",
+	"nousresearch/nous-hermes-2-mixtral-8x7b-dpo",
+	"nousresearch/nous-hermes-llama2-13b",
+	"nvidia/llama-3.1-nemotron-70b-instruct",
+	"nvidia/llama-3.1-nemotron-70b-instruct:free",
+	"openai/chatgpt-4o-latest",
+	"openai/gpt-3.5-turbo",
+	"openai/gpt-3.5-turbo-0125",
+	"openai/gpt-3.5-turbo-0613",
+	"openai/gpt-3.5-turbo-1106",
+	"openai/gpt-3.5-turbo-16k",
+	"openai/gpt-3.5-turbo-instruct",
+	"openai/gpt-4",
+	"openai/gpt-4-0314",
+	"openai/gpt-4-1106-preview",
+	"openai/gpt-4-32k",
+	"openai/gpt-4-32k-0314",
+	"openai/gpt-4-turbo",
+	"openai/gpt-4-turbo-preview",
+	"openai/gpt-4o",
+	"openai/gpt-4o-2024-05-13",
+	"openai/gpt-4o-2024-08-06",
+	"openai/gpt-4o-2024-11-20",
+	"openai/gpt-4o-mini",
+	"openai/gpt-4o-mini-2024-07-18",
+	"openai/gpt-4o:extended",
+	"openai/o1",
+	"openai/o1-mini",
+	"openai/o1-mini-2024-09-12",
+	"openai/o1-preview",
+	"openai/o1-preview-2024-09-12",
+	"openai/o3-mini",
+	"openai/o3-mini-high",
+	"openchat/openchat-7b",
+	"openchat/openchat-7b:free",
+	"openrouter/auto",
+	"perplexity/llama-3.1-sonar-huge-128k-online",
+	"perplexity/llama-3.1-sonar-large-128k-chat",
+	"perplexity/llama-3.1-sonar-large-128k-online",
+	"perplexity/llama-3.1-sonar-small-128k-chat",
+	"perplexity/llama-3.1-sonar-small-128k-online",
+	"perplexity/sonar",
+	"perplexity/sonar-reasoning",
+	"pygmalionai/mythalion-13b",
+	"qwen/qvq-72b-preview",
+	"qwen/qwen-2-72b-instruct",
+	"qwen/qwen-2-7b-instruct",
+	"qwen/qwen-2-7b-instruct:free",
+	"qwen/qwen-2-vl-72b-instruct",
+	"qwen/qwen-2-vl-7b-instruct",
+	"qwen/qwen-2.5-72b-instruct",
+	"qwen/qwen-2.5-7b-instruct",
+	"qwen/qwen-2.5-coder-32b-instruct",
+	"qwen/qwen-max",
+	"qwen/qwen-plus",
+	"qwen/qwen-turbo",
 	"qwen/qwen-vl-plus:free",
+	"qwen/qwen2.5-vl-72b-instruct:free",
+	"qwen/qwq-32b-preview",
+	"raifle/sorcererlm-8x22b",
+	"sao10k/fimbulvetr-11b-v2",
+	"sao10k/l3-euryale-70b",
+	"sao10k/l3-lunaris-8b",
+	"sao10k/l3.1-70b-hanami-x1",
+	"sao10k/l3.1-euryale-70b",
+	"sao10k/l3.3-euryale-70b",
+	"sophosympatheia/midnight-rose-70b",
+	"sophosympatheia/rogue-rose-103b-v0.2:free",
+	"teknium/openhermes-2.5-mistral-7b",
+	"thedrummer/rocinante-12b",
+	"thedrummer/unslopnemo-12b",
+	"undi95/remm-slerp-l2-13b",
+	"undi95/toppy-m-7b",
+	"undi95/toppy-m-7b:free",
+	"x-ai/grok-2-1212",
+	"x-ai/grok-2-vision-1212",
+	"x-ai/grok-beta",
+	"x-ai/grok-vision-beta",
+	"xwin-lm/xwin-lm-70b",
 }

relay/adaptor/openrouter/model.go

@@ -0,0 +1,22 @@
+package openrouter
+
+// RequestProvider customize how your requests are routed using the provider object
+// in the request body for Chat Completions and Completions.
+//
+// https://openrouter.ai/docs/features/provider-routing
+type RequestProvider struct {
+	// Order is list of provider names to try in order (e.g. ["Anthropic", "OpenAI"]). Default: empty
+	Order []string `json:"order,omitempty"`
+	// AllowFallbacks is whether to allow backup providers when the primary is unavailable. Default: true
+	AllowFallbacks bool `json:"allow_fallbacks,omitempty"`
+	// RequireParameters is only use providers that support all parameters in your request. Default: false
+	RequireParameters bool `json:"require_parameters,omitempty"`
+	// DataCollection is control whether to use providers that may store data ("allow" or "deny"). Default: "allow"
+	DataCollection string `json:"data_collection,omitempty" binding:"omitempty,oneof=allow deny"`
+	// Ignore is list of provider names to skip for this request. Default: empty
+	Ignore []string `json:"ignore,omitempty"`
+	// Quantizations is list of quantization levels to filter by (e.g. ["int4", "int8"]). Default: empty
+	Quantizations []string `json:"quantizations,omitempty"`
+	// Sort is sort providers by price or throughput (e.g. "price" or "throughput"). Default: empty
+	Sort string `json:"sort,omitempty" binding:"omitempty,oneof=price throughput latency"`
+}

relay/adaptor/vertexai/claude/adapter.go

@@ -19,6 +19,7 @@ var ModelList = []string{
 	"claude-3-5-sonnet@20240620",
 	"claude-3-5-sonnet-v2@20241022",
 	"claude-3-5-haiku@20241022",
+	"claude-3-7-sonnet@20250219",
 }
 
 const anthropicVersion = "vertex-2023-10-16"
@@ -31,7 +32,11 @@ func (a *Adaptor) ConvertRequest(c *gin.Context, relayMode int, request *model.G
 		return nil, errors.New("request is nil")
 	}
 
-	claudeReq := anthropic.ConvertRequest(*request)
+	claudeReq, err := anthropic.ConvertRequest(c, *request)
+	if err != nil {
+		return nil, errors.Wrap(err, "convert request")
+	}
+
 	req := Request{
 		AnthropicVersion: anthropicVersion,
 		// Model:            claudeReq.Model,

relay/billing/ratio/model.go

@@ -59,6 +59,8 @@ var ModelRatio = map[string]float64{
 	"o1-preview-2024-09-12":   7.5,
 	"o1-mini":                 1.5, // $3.00 / 1M input tokens
 	"o1-mini-2024-09-12":      1.5,
+	"o3-mini":                 1.5, // $3.00 / 1M input tokens
+	"o3-mini-2025-01-31":      1.5,
 	"davinci-002":             1,   // $0.002 / 1K tokens
 	"babbage-002":             0.2, // $0.0004 / 1K tokens
 	"text-ada-001":            0.2,
@@ -96,6 +98,8 @@ var ModelRatio = map[string]float64{
 	"claude-3-5-sonnet-20240620": 3.0 / 1000 * USD,
 	"claude-3-5-sonnet-20241022": 3.0 / 1000 * USD,
 	"claude-3-5-sonnet-latest":   3.0 / 1000 * USD,
+	"claude-3-7-sonnet-20250219": 3.0 / 1000 * USD,
+	"claude-3-7-sonnet-latest":   3.0 / 1000 * USD,
 	"claude-3-opus-20240229":     15.0 / 1000 * USD,
 	// https://cloud.baidu.com/doc/WENXINWORKSHOP/s/hlrk4akp7
 	"ERNIE-4.0-8K":       0.120 * RMB,
@@ -159,91 +163,105 @@ var ModelRatio = map[string]float64{
 	"embedding-2":      0.0005 * RMB,
 	"embedding-3":      0.0005 * RMB,
 	// https://help.aliyun.com/zh/dashscope/developer-reference/tongyi-thousand-questions-metering-and-billing
-	"qwen-turbo":                  1.4286, // ￥0.02 / 1k tokens
-	"qwen-turbo-latest":           1.4286,
-	"qwen-plus":                   1.4286,
-	"qwen-plus-latest":            1.4286,
-	"qwen-max":                    1.4286,
-	"qwen-max-latest":             1.4286,
-	"qwen-max-longcontext":        1.4286,
-	"qwen-vl-max":                 1.4286,
-	"qwen-vl-max-latest":          1.4286,
-	"qwen-vl-plus":                1.4286,
-	"qwen-vl-plus-latest":         1.4286,
-	"qwen-vl-ocr":                 1.4286,
-	"qwen-vl-ocr-latest":          1.4286,
-	"qwen-audio-turbo":            1.4286,
-	"qwen-math-plus":              1.4286,
-	"qwen-math-plus-latest":       1.4286,
-	"qwen-math-turbo":             1.4286,
-	"qwen-math-turbo-latest":      1.4286,
-	"qwen-coder-plus":             1.4286,
-	"qwen-coder-plus-latest":      1.4286,
-	"qwen-coder-turbo":            1.4286,
-	"qwen-coder-turbo-latest":     1.4286,
-	"qwq-32b-preview":             1.4286,
-	"qwen2.5-72b-instruct":        1.4286,
-	"qwen2.5-32b-instruct":        1.4286,
-	"qwen2.5-14b-instruct":        1.4286,
-	"qwen2.5-7b-instruct":         1.4286,
-	"qwen2.5-3b-instruct":         1.4286,
-	"qwen2.5-1.5b-instruct":       1.4286,
-	"qwen2.5-0.5b-instruct":       1.4286,
-	"qwen2-72b-instruct":          1.4286,
-	"qwen2-57b-a14b-instruct":     1.4286,
-	"qwen2-7b-instruct":           1.4286,
-	"qwen2-1.5b-instruct":         1.4286,
-	"qwen2-0.5b-instruct":         1.4286,
-	"qwen1.5-110b-chat":           1.4286,
-	"qwen1.5-72b-chat":            1.4286,
-	"qwen1.5-32b-chat":            1.4286,
-	"qwen1.5-14b-chat":            1.4286,
-	"qwen1.5-7b-chat":             1.4286,
-	"qwen1.5-1.8b-chat":           1.4286,
-	"qwen1.5-0.5b-chat":           1.4286,
-	"qwen-72b-chat":               1.4286,
-	"qwen-14b-chat":               1.4286,
-	"qwen-7b-chat":                1.4286,
-	"qwen-1.8b-chat":              1.4286,
-	"qwen-1.8b-longcontext-chat":  1.4286,
-	"qwen2-vl-7b-instruct":        1.4286,
-	"qwen2-vl-2b-instruct":        1.4286,
-	"qwen-vl-v1":                  1.4286,
-	"qwen-vl-chat-v1":             1.4286,
-	"qwen2-audio-instruct":        1.4286,
-	"qwen-audio-chat":             1.4286,
-	"qwen2.5-math-72b-instruct":   1.4286,
-	"qwen2.5-math-7b-instruct":    1.4286,
-	"qwen2.5-math-1.5b-instruct":  1.4286,
-	"qwen2-math-72b-instruct":     1.4286,
-	"qwen2-math-7b-instruct":      1.4286,
-	"qwen2-math-1.5b-instruct":    1.4286,
-	"qwen2.5-coder-32b-instruct":  1.4286,
-	"qwen2.5-coder-14b-instruct":  1.4286,
-	"qwen2.5-coder-7b-instruct":   1.4286,
-	"qwen2.5-coder-3b-instruct":   1.4286,
-	"qwen2.5-coder-1.5b-instruct": 1.4286,
-	"qwen2.5-coder-0.5b-instruct": 1.4286,
-	"text-embedding-v1":           0.05, // ￥0.0007 / 1k tokens
-	"text-embedding-v3":           0.05,
-	"text-embedding-v2":           0.05,
-	"text-embedding-async-v2":     0.05,
-	"text-embedding-async-v1":     0.05,
-	"ali-stable-diffusion-xl":     8.00,
-	"ali-stable-diffusion-v1.5":   8.00,
-	"wanx-v1":                     8.00,
-	"SparkDesk":                   1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v1.1":              1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v2.1":              1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.1":              1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.1-128K":         1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.5":              1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v3.5-32K":          1.2858, // ￥0.018 / 1k tokens
-	"SparkDesk-v4.0":              1.2858, // ￥0.018 / 1k tokens
-	"360GPT_S2_V9":                0.8572, // ¥0.012 / 1k tokens
-	"embedding-bert-512-v1":       0.0715, // ¥0.001 / 1k tokens
-	"embedding_s1_v1":             0.0715, // ¥0.001 / 1k tokens
-	"semantic_similarity_s1_v1":   0.0715, // ¥0.001 / 1k tokens
+	"qwen-turbo":                    0.0003 * RMB,
+	"qwen-turbo-latest":             0.0003 * RMB,
+	"qwen-plus":                     0.0008 * RMB,
+	"qwen-plus-latest":              0.0008 * RMB,
+	"qwen-max":                      0.0024 * RMB,
+	"qwen-max-latest":               0.0024 * RMB,
+	"qwen-max-longcontext":          0.0005 * RMB,
+	"qwen-vl-max":                   0.003 * RMB,
+	"qwen-vl-max-latest":            0.003 * RMB,
+	"qwen-vl-plus":                  0.0015 * RMB,
+	"qwen-vl-plus-latest":           0.0015 * RMB,
+	"qwen-vl-ocr":                   0.005 * RMB,
+	"qwen-vl-ocr-latest":            0.005 * RMB,
+	"qwen-audio-turbo":              1.4286,
+	"qwen-math-plus":                0.004 * RMB,
+	"qwen-math-plus-latest":         0.004 * RMB,
+	"qwen-math-turbo":               0.002 * RMB,
+	"qwen-math-turbo-latest":        0.002 * RMB,
+	"qwen-coder-plus":               0.0035 * RMB,
+	"qwen-coder-plus-latest":        0.0035 * RMB,
+	"qwen-coder-turbo":              0.002 * RMB,
+	"qwen-coder-turbo-latest":       0.002 * RMB,
+	"qwen-mt-plus":                  0.015 * RMB,
+	"qwen-mt-turbo":                 0.001 * RMB,
+	"qwq-32b-preview":               0.002 * RMB,
+	"qwen2.5-72b-instruct":          0.004 * RMB,
+	"qwen2.5-32b-instruct":          0.03 * RMB,
+	"qwen2.5-14b-instruct":          0.001 * RMB,
+	"qwen2.5-7b-instruct":           0.0005 * RMB,
+	"qwen2.5-3b-instruct":           0.006 * RMB,
+	"qwen2.5-1.5b-instruct":         0.0003 * RMB,
+	"qwen2.5-0.5b-instruct":         0.0003 * RMB,
+	"qwen2-72b-instruct":            0.004 * RMB,
+	"qwen2-57b-a14b-instruct":       0.0035 * RMB,
+	"qwen2-7b-instruct":             0.001 * RMB,
+	"qwen2-1.5b-instruct":           0.001 * RMB,
+	"qwen2-0.5b-instruct":           0.001 * RMB,
+	"qwen1.5-110b-chat":             0.007 * RMB,
+	"qwen1.5-72b-chat":              0.005 * RMB,
+	"qwen1.5-32b-chat":              0.0035 * RMB,
+	"qwen1.5-14b-chat":              0.002 * RMB,
+	"qwen1.5-7b-chat":               0.001 * RMB,
+	"qwen1.5-1.8b-chat":             0.001 * RMB,
+	"qwen1.5-0.5b-chat":             0.001 * RMB,
+	"qwen-72b-chat":                 0.02 * RMB,
+	"qwen-14b-chat":                 0.008 * RMB,
+	"qwen-7b-chat":                  0.006 * RMB,
+	"qwen-1.8b-chat":                0.006 * RMB,
+	"qwen-1.8b-longcontext-chat":    0.006 * RMB,
+	"qvq-72b-preview":               0.012 * RMB,
+	"qwen2.5-vl-72b-instruct":       0.016 * RMB,
+	"qwen2.5-vl-7b-instruct":        0.002 * RMB,
+	"qwen2.5-vl-3b-instruct":        0.0012 * RMB,
+	"qwen2-vl-7b-instruct":          0.016 * RMB,
+	"qwen2-vl-2b-instruct":          0.002 * RMB,
+	"qwen-vl-v1":                    0.002 * RMB,
+	"qwen-vl-chat-v1":               0.002 * RMB,
+	"qwen2-audio-instruct":          0.002 * RMB,
+	"qwen-audio-chat":               0.002 * RMB,
+	"qwen2.5-math-72b-instruct":     0.004 * RMB,
+	"qwen2.5-math-7b-instruct":      0.001 * RMB,
+	"qwen2.5-math-1.5b-instruct":    0.001 * RMB,
+	"qwen2-math-72b-instruct":       0.004 * RMB,
+	"qwen2-math-7b-instruct":        0.001 * RMB,
+	"qwen2-math-1.5b-instruct":      0.001 * RMB,
+	"qwen2.5-coder-32b-instruct":    0.002 * RMB,
+	"qwen2.5-coder-14b-instruct":    0.002 * RMB,
+	"qwen2.5-coder-7b-instruct":     0.001 * RMB,
+	"qwen2.5-coder-3b-instruct":     0.001 * RMB,
+	"qwen2.5-coder-1.5b-instruct":   0.001 * RMB,
+	"qwen2.5-coder-0.5b-instruct":   0.001 * RMB,
+	"text-embedding-v1":             0.0007 * RMB, // ￥0.0007 / 1k tokens
+	"text-embedding-v3":             0.0007 * RMB,
+	"text-embedding-v2":             0.0007 * RMB,
+	"text-embedding-async-v2":       0.0007 * RMB,
+	"text-embedding-async-v1":       0.0007 * RMB,
+	"ali-stable-diffusion-xl":       8.00,
+	"ali-stable-diffusion-v1.5":     8.00,
+	"wanx-v1":                       8.00,
+	"deepseek-r1":                   0.002 * RMB,
+	"deepseek-v3":                   0.001 * RMB,
+	"deepseek-r1-distill-qwen-1.5b": 0.001 * RMB,
+	"deepseek-r1-distill-qwen-7b":   0.0005 * RMB,
+	"deepseek-r1-distill-qwen-14b":  0.001 * RMB,
+	"deepseek-r1-distill-qwen-32b":  0.002 * RMB,
+	"deepseek-r1-distill-llama-8b":  0.0005 * RMB,
+	"deepseek-r1-distill-llama-70b": 0.004 * RMB,
+	"SparkDesk":                     1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v1.1":                1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v2.1":                1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.1":                1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.1-128K":           1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.5":                1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v3.5-32K":            1.2858, // ￥0.018 / 1k tokens
+	"SparkDesk-v4.0":                1.2858, // ￥0.018 / 1k tokens
+	"360GPT_S2_V9":                  0.8572, // ¥0.012 / 1k tokens
+	"embedding-bert-512-v1":         0.0715, // ¥0.001 / 1k tokens
+	"embedding_s1_v1":               0.0715, // ¥0.001 / 1k tokens
+	"semantic_similarity_s1_v1":     0.0715, // ¥0.001 / 1k tokens
 	// https://cloud.tencent.com/document/product/1729/97731#e0e6be58-60c8-469f-bdeb-6c264ce3b4d0
 	"hunyuan-turbo":             0.015 * RMB,
 	"hunyuan-large":             0.004 * RMB,
@@ -371,6 +389,238 @@ var ModelRatio = map[string]float64{
 	"mistralai/mistral-7b-instruct-v0.2":        0.050 * USD,
 	"mistralai/mistral-7b-v0.1":                 0.050 * USD,
 	"mistralai/mixtral-8x7b-instruct-v0.1":      0.300 * USD,
+	//https://openrouter.ai/models
+	"01-ai/yi-large":                                  1.5,
+	"aetherwiing/mn-starcannon-12b":                   0.6,
+	"ai21/jamba-1-5-large":                            4.0,
+	"ai21/jamba-1-5-mini":                             0.2,
+	"ai21/jamba-instruct":                             0.35,
+	"aion-labs/aion-1.0":                              6.0,
+	"aion-labs/aion-1.0-mini":                         1.2,
+	"aion-labs/aion-rp-llama-3.1-8b":                  0.1,
+	"allenai/llama-3.1-tulu-3-405b":                   5.0,
+	"alpindale/goliath-120b":                          4.6875,
+	"alpindale/magnum-72b":                            1.125,
+	"amazon/nova-lite-v1":                             0.12,
+	"amazon/nova-micro-v1":                            0.07,
+	"amazon/nova-pro-v1":                              1.6,
+	"anthracite-org/magnum-v2-72b":                    1.5,
+	"anthracite-org/magnum-v4-72b":                    1.125,
+	"anthropic/claude-2":                              12.0,
+	"anthropic/claude-2.0":                            12.0,
+	"anthropic/claude-2.0:beta":                       12.0,
+	"anthropic/claude-2.1":                            12.0,
+	"anthropic/claude-2.1:beta":                       12.0,
+	"anthropic/claude-2:beta":                         12.0,
+	"anthropic/claude-3-haiku":                        0.625,
+	"anthropic/claude-3-haiku:beta":                   0.625,
+	"anthropic/claude-3-opus":                         37.5,
+	"anthropic/claude-3-opus:beta":                    37.5,
+	"anthropic/claude-3-sonnet":                       7.5,
+	"anthropic/claude-3-sonnet:beta":                  7.5,
+	"anthropic/claude-3.5-haiku":                      2.0,
+	"anthropic/claude-3.5-haiku-20241022":             2.0,
+	"anthropic/claude-3.5-haiku-20241022:beta":        2.0,
+	"anthropic/claude-3.5-haiku:beta":                 2.0,
+	"anthropic/claude-3.5-sonnet":                     7.5,
+	"anthropic/claude-3.5-sonnet-20240620":            7.5,
+	"anthropic/claude-3.5-sonnet-20240620:beta":       7.5,
+	"anthropic/claude-3.5-sonnet:beta":                7.5,
+	"cognitivecomputations/dolphin-mixtral-8x22b":     0.45,
+	"cognitivecomputations/dolphin-mixtral-8x7b":      0.25,
+	"cohere/command":                                  0.95,
+	"cohere/command-r":                                0.7125,
+	"cohere/command-r-03-2024":                        0.7125,
+	"cohere/command-r-08-2024":                        0.285,
+	"cohere/command-r-plus":                           7.125,
+	"cohere/command-r-plus-04-2024":                   7.125,
+	"cohere/command-r-plus-08-2024":                   4.75,
+	"cohere/command-r7b-12-2024":                      0.075,
+	"databricks/dbrx-instruct":                        0.6,
+	"deepseek/deepseek-chat":                          0.445,
+	"deepseek/deepseek-chat-v2.5":                     1.0,
+	"deepseek/deepseek-chat:free":                     0.0,
+	"deepseek/deepseek-r1":                            1.2,
+	"deepseek/deepseek-r1-distill-llama-70b":          0.345,
+	"deepseek/deepseek-r1-distill-llama-70b:free":     0.0,
+	"deepseek/deepseek-r1-distill-llama-8b":           0.02,
+	"deepseek/deepseek-r1-distill-qwen-1.5b":          0.09,
+	"deepseek/deepseek-r1-distill-qwen-14b":           0.075,
+	"deepseek/deepseek-r1-distill-qwen-32b":           0.09,
+	"deepseek/deepseek-r1:free":                       0.0,
+	"eva-unit-01/eva-llama-3.33-70b":                  3.0,
+	"eva-unit-01/eva-qwen-2.5-32b":                    1.7,
+	"eva-unit-01/eva-qwen-2.5-72b":                    3.0,
+	"google/gemini-2.0-flash-001":                     0.2,
+	"google/gemini-2.0-flash-exp:free":                0.0,
+	"google/gemini-2.0-flash-lite-preview-02-05:free": 0.0,
+	"google/gemini-2.0-flash-thinking-exp-1219:free":  0.0,
+	"google/gemini-2.0-flash-thinking-exp:free":       0.0,
+	"google/gemini-2.0-pro-exp-02-05:free":            0.0,
+	"google/gemini-exp-1206:free":                     0.0,
+	"google/gemini-flash-1.5":                         0.15,
+	"google/gemini-flash-1.5-8b":                      0.075,
+	"google/gemini-flash-1.5-8b-exp":                  0.0,
+	"google/gemini-pro":                               0.75,
+	"google/gemini-pro-1.5":                           2.5,
+	"google/gemini-pro-vision":                        0.75,
+	"google/gemma-2-27b-it":                           0.135,
+	"google/gemma-2-9b-it":                            0.03,
+	"google/gemma-2-9b-it:free":                       0.0,
+	"google/gemma-7b-it":                              0.075,
+	"google/learnlm-1.5-pro-experimental:free":        0.0,
+	"google/palm-2-chat-bison":                        1.0,
+	"google/palm-2-chat-bison-32k":                    1.0,
+	"google/palm-2-codechat-bison":                    1.0,
+	"google/palm-2-codechat-bison-32k":                1.0,
+	"gryphe/mythomax-l2-13b":                          0.0325,
+	"gryphe/mythomax-l2-13b:free":                     0.0,
+	"huggingfaceh4/zephyr-7b-beta:free":               0.0,
+	"infermatic/mn-inferor-12b":                       0.6,
+	"inflection/inflection-3-pi":                      5.0,
+	"inflection/inflection-3-productivity":            5.0,
+	"jondurbin/airoboros-l2-70b":                      0.25,
+	"liquid/lfm-3b":                                   0.01,
+	"liquid/lfm-40b":                                  0.075,
+	"liquid/lfm-7b":                                   0.005,
+	"mancer/weaver":                                   1.125,
+	"meta-llama/llama-2-13b-chat":                     0.11,
+	"meta-llama/llama-2-70b-chat":                     0.45,
+	"meta-llama/llama-3-70b-instruct":                 0.2,
+	"meta-llama/llama-3-8b-instruct":                  0.03,
+	"meta-llama/llama-3-8b-instruct:free":             0.0,
+	"meta-llama/llama-3.1-405b":                       1.0,
+	"meta-llama/llama-3.1-405b-instruct":              0.4,
+	"meta-llama/llama-3.1-70b-instruct":               0.15,
+	"meta-llama/llama-3.1-8b-instruct":                0.025,
+	"meta-llama/llama-3.2-11b-vision-instruct":        0.0275,
+	"meta-llama/llama-3.2-11b-vision-instruct:free":   0.0,
+	"meta-llama/llama-3.2-1b-instruct":                0.005,
+	"meta-llama/llama-3.2-3b-instruct":                0.0125,
+	"meta-llama/llama-3.2-90b-vision-instruct":        0.8,
+	"meta-llama/llama-3.3-70b-instruct":               0.15,
+	"meta-llama/llama-3.3-70b-instruct:free":          0.0,
+	"meta-llama/llama-guard-2-8b":                     0.1,
+	"microsoft/phi-3-medium-128k-instruct":            0.5,
+	"microsoft/phi-3-medium-128k-instruct:free":       0.0,
+	"microsoft/phi-3-mini-128k-instruct":              0.05,
+	"microsoft/phi-3-mini-128k-instruct:free":         0.0,
+	"microsoft/phi-3.5-mini-128k-instruct":            0.05,
+	"microsoft/phi-4":                                 0.07,
+	"microsoft/wizardlm-2-7b":                         0.035,
+	"microsoft/wizardlm-2-8x22b":                      0.25,
+	"minimax/minimax-01":                              0.55,
+	"mistralai/codestral-2501":                        0.45,
+	"mistralai/codestral-mamba":                       0.125,
+	"mistralai/ministral-3b":                          0.02,
+	"mistralai/ministral-8b":                          0.05,
+	"mistralai/mistral-7b-instruct":                   0.0275,
+	"mistralai/mistral-7b-instruct-v0.1":              0.1,
+	"mistralai/mistral-7b-instruct-v0.3":              0.0275,
+	"mistralai/mistral-7b-instruct:free":              0.0,
+	"mistralai/mistral-large":                         3.0,
+	"mistralai/mistral-large-2407":                    3.0,
+	"mistralai/mistral-large-2411":                    3.0,
+	"mistralai/mistral-medium":                        4.05,
+	"mistralai/mistral-nemo":                          0.04,
+	"mistralai/mistral-nemo:free":                     0.0,
+	"mistralai/mistral-small":                         0.3,
+	"mistralai/mistral-small-24b-instruct-2501":       0.07,
+	"mistralai/mistral-small-24b-instruct-2501:free":  0.0,
+	"mistralai/mistral-tiny":                          0.125,
+	"mistralai/mixtral-8x22b-instruct":                0.45,
+	"mistralai/mixtral-8x7b":                          0.3,
+	"mistralai/mixtral-8x7b-instruct":                 0.12,
+	"mistralai/pixtral-12b":                           0.05,
+	"mistralai/pixtral-large-2411":                    3.0,
+	"neversleep/llama-3-lumimaid-70b":                 2.25,
+	"neversleep/llama-3-lumimaid-8b":                  0.5625,
+	"neversleep/llama-3-lumimaid-8b:extended":         0.5625,
+	"neversleep/llama-3.1-lumimaid-70b":               2.25,
+	"neversleep/llama-3.1-lumimaid-8b":                0.5625,
+	"neversleep/noromaid-20b":                         1.125,
+	"nothingiisreal/mn-celeste-12b":                   0.6,
+	"nousresearch/hermes-2-pro-llama-3-8b":            0.02,
+	"nousresearch/hermes-3-llama-3.1-405b":            0.4,
+	"nousresearch/hermes-3-llama-3.1-70b":             0.15,
+	"nousresearch/nous-hermes-2-mixtral-8x7b-dpo":     0.3,
+	"nousresearch/nous-hermes-llama2-13b":             0.085,
+	"nvidia/llama-3.1-nemotron-70b-instruct":          0.15,
+	"nvidia/llama-3.1-nemotron-70b-instruct:free":     0.0,
+	"openai/chatgpt-4o-latest":                        7.5,
+	"openai/gpt-3.5-turbo":                            0.75,
+	"openai/gpt-3.5-turbo-0125":                       0.75,
+	"openai/gpt-3.5-turbo-0613":                       1.0,
+	"openai/gpt-3.5-turbo-1106":                       1.0,
+	"openai/gpt-3.5-turbo-16k":                        2.0,
+	"openai/gpt-3.5-turbo-instruct":                   1.0,
+	"openai/gpt-4":                                    30.0,
+	"openai/gpt-4-0314":                               30.0,
+	"openai/gpt-4-1106-preview":                       15.0,
+	"openai/gpt-4-32k":                                60.0,
+	"openai/gpt-4-32k-0314":                           60.0,
+	"openai/gpt-4-turbo":                              15.0,
+	"openai/gpt-4-turbo-preview":                      15.0,
+	"openai/gpt-4o":                                   5.0,
+	"openai/gpt-4o-2024-05-13":                        7.5,
+	"openai/gpt-4o-2024-08-06":                        5.0,
+	"openai/gpt-4o-2024-11-20":                        5.0,
+	"openai/gpt-4o-mini":                              0.3,
+	"openai/gpt-4o-mini-2024-07-18":                   0.3,
+	"openai/gpt-4o:extended":                          9.0,
+	"openai/o1":                                       30.0,
+	"openai/o1-mini":                                  2.2,
+	"openai/o1-mini-2024-09-12":                       2.2,
+	"openai/o1-preview":                               30.0,
+	"openai/o1-preview-2024-09-12":                    30.0,
+	"openai/o3-mini":                                  2.2,
+	"openai/o3-mini-high":                             2.2,
+	"openchat/openchat-7b":                            0.0275,
+	"openchat/openchat-7b:free":                       0.0,
+	"openrouter/auto":                                 -500000.0,
+	"perplexity/llama-3.1-sonar-huge-128k-online":     2.5,
+	"perplexity/llama-3.1-sonar-large-128k-chat":      0.5,
+	"perplexity/llama-3.1-sonar-large-128k-online":    0.5,
+	"perplexity/llama-3.1-sonar-small-128k-chat":      0.1,
+	"perplexity/llama-3.1-sonar-small-128k-online":    0.1,
+	"perplexity/sonar":                                0.5,
+	"perplexity/sonar-reasoning":                      2.5,
+	"pygmalionai/mythalion-13b":                       0.6,
+	"qwen/qvq-72b-preview":                            0.25,
+	"qwen/qwen-2-72b-instruct":                        0.45,
+	"qwen/qwen-2-7b-instruct":                         0.027,
+	"qwen/qwen-2-7b-instruct:free":                    0.0,
+	"qwen/qwen-2-vl-72b-instruct":                     0.2,
+	"qwen/qwen-2-vl-7b-instruct":                      0.05,
+	"qwen/qwen-2.5-72b-instruct":                      0.2,
+	"qwen/qwen-2.5-7b-instruct":                       0.025,
+	"qwen/qwen-2.5-coder-32b-instruct":                0.08,
+	"qwen/qwen-max":                                   3.2,
+	"qwen/qwen-plus":                                  0.6,
+	"qwen/qwen-turbo":                                 0.1,
+	"qwen/qwen-vl-plus:free":                          0.0,
+	"qwen/qwen2.5-vl-72b-instruct:free":               0.0,
+	"qwen/qwq-32b-preview":                            0.09,
+	"raifle/sorcererlm-8x22b":                         2.25,
+	"sao10k/fimbulvetr-11b-v2":                        0.6,
+	"sao10k/l3-euryale-70b":                           0.4,
+	"sao10k/l3-lunaris-8b":                            0.03,
+	"sao10k/l3.1-70b-hanami-x1":                       1.5,
+	"sao10k/l3.1-euryale-70b":                         0.4,
+	"sao10k/l3.3-euryale-70b":                         0.4,
+	"sophosympatheia/midnight-rose-70b":               0.4,
+	"sophosympatheia/rogue-rose-103b-v0.2:free":       0.0,
+	"teknium/openhermes-2.5-mistral-7b":               0.085,
+	"thedrummer/rocinante-12b":                        0.25,
+	"thedrummer/unslopnemo-12b":                       0.25,
+	"undi95/remm-slerp-l2-13b":                        0.6,
+	"undi95/toppy-m-7b":                               0.035,
+	"undi95/toppy-m-7b:free":                          0.0,
+	"x-ai/grok-2-1212":                                5.0,
+	"x-ai/grok-2-vision-1212":                         5.0,
+	"x-ai/grok-beta":                                  7.5,
+	"x-ai/grok-vision-beta":                           7.5,
+	"xwin-lm/xwin-lm-70b":                             1.875,
 }
 
 var CompletionRatio = map[string]float64{

relay/controller/helper.go

@@ -102,6 +102,9 @@ func postConsumeQuota(ctx context.Context, usage *relaymodel.Usage, meta *meta.M
 	var quota int64
 	completionRatio := billingratio.GetCompletionRatio(textRequest.Model, meta.ChannelType)
 	promptTokens := usage.PromptTokens
+	// It appears that DeepSeek's official service automatically merges ReasoningTokens into CompletionTokens,
+	// but the behavior of third-party providers may differ, so for now we do not add them manually.
+	// completionTokens := usage.CompletionTokens + usage.CompletionTokensDetails.ReasoningTokens
 	completionTokens := usage.CompletionTokens
 	quota = int64(math.Ceil((float64(promptTokens) + float64(completionTokens)*completionRatio) * ratio))
 	if ratio != 0 && quota <= 0 {

relay/model/general.go

@@ -1,5 +1,7 @@
 package model
 
+import "github.com/songquanpeng/one-api/relay/adaptor/openrouter"
+
 type ResponseFormat struct {
 	Type       string      `json:"type,omitempty"`
 	JsonSchema *JSONSchema `json:"json_schema,omitempty"`
@@ -66,6 +68,21 @@ type GeneralOpenAIRequest struct {
 	// Others
 	Instruction string `json:"instruction,omitempty"`
 	NumCtx      int    `json:"num_ctx,omitempty"`
+	// -------------------------------------
+	// Openrouter
+	// -------------------------------------
+	Provider         *openrouter.RequestProvider `json:"provider,omitempty"`
+	IncludeReasoning *bool                       `json:"include_reasoning,omitempty"`
+	// -------------------------------------
+	// Anthropic
+	// -------------------------------------
+	Thinking *Thinking `json:"thinking,omitempty"`
+}
+
+// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#implementing-extended-thinking
+type Thinking struct {
+	Type         string `json:"type"`
+	BudgetTokens int    `json:"budget_tokens" binding:"omitempty,min=1024"`
 }
 
 func (r GeneralOpenAIRequest) ParseInput() []string {

relay/model/message.go

@@ -1,12 +1,35 @@
 package model
 
 type Message struct {
-	Role             string  `json:"role,omitempty"`
-	Content          any     `json:"content,omitempty"`
-	ReasoningContent any     `json:"reasoning_content,omitempty"`
-	Name             *string `json:"name,omitempty"`
-	ToolCalls        []Tool  `json:"tool_calls,omitempty"`
-	ToolCallId       string  `json:"tool_call_id,omitempty"`
+	Role string `json:"role,omitempty"`
+	// Content is a string or a list of objects
+	Content    any           `json:"content,omitempty"`
+	Name       *string       `json:"name,omitempty"`
+	ToolCalls  []Tool        `json:"tool_calls,omitempty"`
+	ToolCallId string        `json:"tool_call_id,omitempty"`
+	Audio      *messageAudio `json:"audio,omitempty"`
+	// -------------------------------------
+	// Deepseek 专有的一些字段
+	// https://api-docs.deepseek.com/api/create-chat-completion
+	// -------------------------------------
+	// Prefix forces the model to begin its answer with the supplied prefix in the assistant message.
+	// To enable this feature, set base_url to "https://api.deepseek.com/beta".
+	Prefix *bool `json:"prefix,omitempty"` // ReasoningContent is Used for the deepseek-reasoner model in the Chat
+	// Prefix Completion feature as the input for the CoT in the last assistant message.
+	// When using this feature, the prefix parameter must be set to true.
+	ReasoningContent *string `json:"reasoning_content,omitempty"`
+	// -------------------------------------
+	// Openrouter
+	// -------------------------------------
+	Reasoning *string `json:"reasoning,omitempty"`
+	Refusal   *bool   `json:"refusal,omitempty"`
+}
+
+type messageAudio struct {
+	Id         string `json:"id"`
+	Data       string `json:"data,omitempty"`
+	ExpiredAt  int    `json:"expired_at,omitempty"`
+	Transcript string `json:"transcript,omitempty"`
 }
 
 func (m Message) IsStringContent() bool {

relay/model/misc.go

@@ -4,14 +4,12 @@ type Usage struct {
 	PromptTokens     int `json:"prompt_tokens"`
 	CompletionTokens int `json:"completion_tokens"`
 	TotalTokens      int `json:"total_tokens"`
-
-	CompletionTokensDetails *CompletionTokensDetails `json:"completion_tokens_details,omitempty"`
-}
-
-type CompletionTokensDetails struct {
-	ReasoningTokens          int `json:"reasoning_tokens"`
-	AcceptedPredictionTokens int `json:"accepted_prediction_tokens"`
-	RejectedPredictionTokens int `json:"rejected_prediction_tokens"`
+	// PromptTokensDetails may be empty for some models
+	PromptTokensDetails *usagePromptTokensDetails `gorm:"-" json:"prompt_tokens_details,omitempty"`
+	// CompletionTokensDetails may be empty for some models
+	CompletionTokensDetails *usageCompletionTokensDetails `gorm:"-" json:"completion_tokens_details,omitempty"`
+	ServiceTier             string                        `gorm:"-" json:"service_tier,omitempty"`
+	SystemFingerprint       string                        `gorm:"-" json:"system_fingerprint,omitempty"`
 }
 
 type Error struct {
@@ -25,3 +23,20 @@ type ErrorWithStatusCode struct {
 	Error
 	StatusCode int `json:"status_code"`
 }
+
+type usagePromptTokensDetails struct {
+	CachedTokens int `json:"cached_tokens"`
+	AudioTokens  int `json:"audio_tokens"`
+	// TextTokens could be zero for pure text chats
+	TextTokens  int `json:"text_tokens"`
+	ImageTokens int `json:"image_tokens"`
+}
+
+type usageCompletionTokensDetails struct {
+	ReasoningTokens          int `json:"reasoning_tokens"`
+	AudioTokens              int `json:"audio_tokens"`
+	AcceptedPredictionTokens int `json:"accepted_prediction_tokens"`
+	RejectedPredictionTokens int `json:"rejected_prediction_tokens"`
+	// TextTokens could be zero for pure text chats
+	TextTokens int `json:"text_tokens"`
+}