diff --git a/dto/openai_request.go b/dto/openai_request.go
index e491812a63ed..6c358f5b5fdd 100644
--- a/dto/openai_request.go
+++ b/dto/openai_request.go
@@ -18,42 +18,42 @@ type FormatJsonSchema struct {
 }
 
 type GeneralOpenAIRequest struct {
-	Model               string         `json:"model,omitempty"`
-	Messages            []Message      `json:"messages,omitempty"`
-	Prompt              any            `json:"prompt,omitempty"`
-	Prefix              any            `json:"prefix,omitempty"`
-	Suffix              any            `json:"suffix,omitempty"`
-	Stream              bool           `json:"stream,omitempty"`
-	StreamOptions       *StreamOptions `json:"stream_options,omitempty"`
-	MaxTokens           uint           `json:"max_tokens,omitempty"`
-	MaxCompletionTokens uint           `json:"max_completion_tokens,omitempty"`
-	ReasoningEffort     string         `json:"reasoning_effort,omitempty"`
-	//Reasoning           json.RawMessage   `json:"reasoning,omitempty"`
-	Temperature      *float64          `json:"temperature,omitempty"`
-	TopP             float64           `json:"top_p,omitempty"`
-	TopK             int               `json:"top_k,omitempty"`
-	Stop             any               `json:"stop,omitempty"`
-	N                int               `json:"n,omitempty"`
-	Input            any               `json:"input,omitempty"`
-	Instruction      string            `json:"instruction,omitempty"`
-	Size             string            `json:"size,omitempty"`
-	Functions        any               `json:"functions,omitempty"`
-	FrequencyPenalty float64           `json:"frequency_penalty,omitempty"`
-	PresencePenalty  float64           `json:"presence_penalty,omitempty"`
-	ResponseFormat   *ResponseFormat   `json:"response_format,omitempty"`
-	EncodingFormat   any               `json:"encoding_format,omitempty"`
-	Seed             float64           `json:"seed,omitempty"`
-	Tools            []ToolCallRequest `json:"tools,omitempty"`
-	ToolChoice       any               `json:"tool_choice,omitempty"`
-	User             string            `json:"user,omitempty"`
-	LogProbs         bool              `json:"logprobs,omitempty"`
-	TopLogProbs      int               `json:"top_logprobs,omitempty"`
-	Dimensions       int               `json:"dimensions,omitempty"`
-	Modalities       any               `json:"modalities,omitempty"`
-	Audio            any               `json:"audio,omitempty"`
-	EnableThinking   any               `json:"enable_thinking,omitempty"` // ali
-	ExtraBody        any               `json:"extra_body,omitempty"`
-	WebSearchOptions *WebSearchOptions `json:"web_search_options,omitempty"`
+	Model               string            `json:"model,omitempty"`
+	Messages            []Message         `json:"messages,omitempty"`
+	Prompt              any               `json:"prompt,omitempty"`
+	Prefix              any               `json:"prefix,omitempty"`
+	Suffix              any               `json:"suffix,omitempty"`
+	Stream              bool              `json:"stream,omitempty"`
+	StreamOptions       *StreamOptions    `json:"stream_options,omitempty"`
+	MaxTokens           uint              `json:"max_tokens,omitempty"`
+	MaxCompletionTokens uint              `json:"max_completion_tokens,omitempty"`
+	ReasoningEffort     string            `json:"reasoning_effort,omitempty"`
+	Reasoning           ReasoningRequest  `json:"reasoning,omitempty"`
+	Temperature         *float64          `json:"temperature,omitempty"`
+	TopP                float64           `json:"top_p,omitempty"`
+	TopK                int               `json:"top_k,omitempty"`
+	Stop                any               `json:"stop,omitempty"`
+	N                   int               `json:"n,omitempty"`
+	Input               any               `json:"input,omitempty"`
+	Instruction         string            `json:"instruction,omitempty"`
+	Size                string            `json:"size,omitempty"`
+	Functions           any               `json:"functions,omitempty"`
+	FrequencyPenalty    float64           `json:"frequency_penalty,omitempty"`
+	PresencePenalty     float64           `json:"presence_penalty,omitempty"`
+	ResponseFormat      *ResponseFormat   `json:"response_format,omitempty"`
+	EncodingFormat      any               `json:"encoding_format,omitempty"`
+	Seed                float64           `json:"seed,omitempty"`
+	Tools               []ToolCallRequest `json:"tools,omitempty"`
+	ToolChoice          any               `json:"tool_choice,omitempty"`
+	User                string            `json:"user,omitempty"`
+	LogProbs            bool              `json:"logprobs,omitempty"`
+	TopLogProbs         int               `json:"top_logprobs,omitempty"`
+	Dimensions          int               `json:"dimensions,omitempty"`
+	Modalities          any               `json:"modalities,omitempty"`
+	Audio               any               `json:"audio,omitempty"`
+	EnableThinking      any               `json:"enable_thinking,omitempty"` // ali
+	ExtraBody           any               `json:"extra_body,omitempty"`
+	WebSearchOptions    *WebSearchOptions `json:"web_search_options,omitempty"`
 }
 
 type ToolCallRequest struct {
@@ -69,6 +69,11 @@ type FunctionRequest struct {
 	Arguments   string `json:"arguments,omitempty"`
 }
 
+type ReasoningRequest struct {
+	Effort            string `json:"effort,omitempty"`
+	MaxThinkingTokens int    `json:"max_tokens,omitempty"`
+}
+
 type StreamOptions struct {
 	IncludeUsage bool `json:"include_usage,omitempty"`
 }
diff --git a/relay/channel/claude/relay-claude.go b/relay/channel/claude/relay-claude.go
index 95e7c4bee4d9..395862b5a42b 100644
--- a/relay/channel/claude/relay-claude.go
+++ b/relay/channel/claude/relay-claude.go
@@ -102,8 +102,15 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*dto.Cla
 		claudeRequest.MaxTokens = uint(model_setting.GetClaudeSettings().GetDefaultMaxTokens(textRequest.Model))
 	}
 
-	if model_setting.GetClaudeSettings().ThinkingAdapterEnabled &&
-		strings.HasSuffix(textRequest.Model, "-thinking") {
+	if (model_setting.GetClaudeSettings().ThinkingAdapterEnabled &&
+		strings.HasSuffix(textRequest.Model, "-thinking")) || textRequest.Reasoning.MaxThinkingTokens > 0 {
+
+		var budgetTokens int
+		if textRequest.Reasoning.MaxThinkingTokens > 0 {
+			budgetTokens = textRequest.Reasoning.MaxThinkingTokens
+		} else {
+			budgetTokens = int(float64(claudeRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage)
+		}
 
 		// 因为BudgetTokens 必须大于1024
 		if claudeRequest.MaxTokens < 1280 {
@@ -113,7 +120,7 @@ func RequestOpenAI2ClaudeMessage(textRequest dto.GeneralOpenAIRequest) (*dto.Cla
 		// BudgetTokens 为 max_tokens 的 80%
 		claudeRequest.Thinking = &dto.Thinking{
 			Type:         "enabled",
-			BudgetTokens: int(float64(claudeRequest.MaxTokens) * model_setting.GetClaudeSettings().ThinkingAdapterBudgetTokensPercentage),
+			BudgetTokens: budgetTokens,
 		}
 		// TODO: 临时处理
 		// https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#important-considerations-when-using-extended-thinking