fix(executor): use upstream model for thinking config and payload translation

2026-02-03 13:00:52 +08:00 · 2025-12-30 17:49:44 +08:00
parent 857c880f99
commit b055e00c1a
8 changed files with 255 additions and 162 deletions
--- a/internal/runtime/executor/aistudio_executor.go
+++ b/internal/runtime/executor/aistudio_executor.go
@@ -55,11 +55,17 @@ func (e *AIStudioExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth,
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

-	translatedReq, body, err := e.translateRequest(req, opts, false)
+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
+	translatedReq, body, err := e.translateRequest(req, opts, false, upstreamModel)
 	if err != nil {
 		return resp, err
 	}
-	endpoint := e.buildEndpoint(req.Model, body.action, opts.Alt)
+
+	endpoint := e.buildEndpoint(upstreamModel, body.action, opts.Alt)
 	wsReq := &wsrelay.HTTPRequest{
 		Method:  http.MethodPost,
 		URL:     endpoint,
@@ -109,11 +115,17 @@ func (e *AIStudioExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

-	translatedReq, body, err := e.translateRequest(req, opts, true)
+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
+	translatedReq, body, err := e.translateRequest(req, opts, true, upstreamModel)
 	if err != nil {
 		return nil, err
 	}
-	endpoint := e.buildEndpoint(req.Model, body.action, opts.Alt)
+
+	endpoint := e.buildEndpoint(upstreamModel, body.action, opts.Alt)
 	wsReq := &wsrelay.HTTPRequest{
 		Method:  http.MethodPost,
 		URL:     endpoint,
@@ -254,7 +266,12 @@ func (e *AIStudioExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth

 // CountTokens counts tokens for the given request using the AI Studio API.
 func (e *AIStudioExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
-	_, body, err := e.translateRequest(req, opts, false)
+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
+	_, body, err := e.translateRequest(req, opts, false, upstreamModel)
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}
@@ -263,7 +280,7 @@ func (e *AIStudioExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.A
 	body.payload, _ = sjson.DeleteBytes(body.payload, "tools")
 	body.payload, _ = sjson.DeleteBytes(body.payload, "safetySettings")

-	endpoint := e.buildEndpoint(req.Model, "countTokens", "")
+	endpoint := e.buildEndpoint(upstreamModel, "countTokens", "")
 	wsReq := &wsrelay.HTTPRequest{
 		Method:  http.MethodPost,
 		URL:     endpoint,
@@ -318,18 +335,23 @@ type translatedPayload struct {
 	toFormat sdktranslator.Format
 }

-func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts cliproxyexecutor.Options, stream bool) ([]byte, translatedPayload, error) {
+func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts cliproxyexecutor.Options, stream bool, upstreamModel string) ([]byte, translatedPayload, error) {
+	model := strings.TrimSpace(upstreamModel)
+	if model == "" {
+		model = strings.TrimSpace(req.Model)
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	payload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
-	payload = ApplyThinkingMetadata(payload, req.Metadata, req.Model)
-	payload = util.ApplyGemini3ThinkingLevelFromMetadata(req.Model, req.Metadata, payload)
-	payload = util.ApplyDefaultThinkingIfNeeded(req.Model, payload)
-	payload = util.ConvertThinkingLevelToBudget(payload, req.Model, true)
-	payload = util.NormalizeGeminiThinkingBudget(req.Model, payload, true)
-	payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
-	payload = fixGeminiImageAspectRatio(req.Model, payload)
-	payload = applyPayloadConfig(e.cfg, req.Model, payload)
+	payload := sdktranslator.TranslateRequest(from, to, model, bytes.Clone(req.Payload), stream)
+	payload = ApplyThinkingMetadata(payload, req.Metadata, model)
+	payload = util.ApplyGemini3ThinkingLevelFromMetadata(model, req.Metadata, payload)
+	payload = util.ApplyDefaultThinkingIfNeeded(model, payload)
+	payload = util.ConvertThinkingLevelToBudget(payload, model, true)
+	payload = util.NormalizeGeminiThinkingBudget(model, payload, true)
+	payload = util.StripThinkingConfigIfUnsupported(model, payload)
+	payload = fixGeminiImageAspectRatio(model, payload)
+	payload = applyPayloadConfig(e.cfg, model, payload)
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.maxOutputTokens")
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseMimeType")
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseJsonSchema")
--- a/internal/runtime/executor/antigravity_executor.go
+++ b/internal/runtime/executor/antigravity_executor.go
@@ -98,13 +98,13 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("antigravity")
-	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	translated := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)

-	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
-	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, translated)
-	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
-	translated = normalizeAntigravityThinking(req.Model, translated, isClaude)
-	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, "antigravity", "request", translated)
+	translated = applyThinkingMetadataCLI(translated, req.Metadata, upstreamModel)
+	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, translated)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, translated)
+	translated = normalizeAntigravityThinking(upstreamModel, translated, isClaude)
+	translated = applyPayloadConfigWithRoot(e.cfg, upstreamModel, "antigravity", "request", translated)

 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
@@ -191,20 +191,20 @@ func (e *AntigravityExecutor) executeClaudeNonStream(ctx context.Context, auth *
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

-	from := opts.SourceFormat
-	to := sdktranslator.FromString("antigravity")
-	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
 	if upstreamModel == "" {
 		upstreamModel = req.Model
 	}

-	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
-	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, translated)
-	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
-	translated = normalizeAntigravityThinking(req.Model, translated, true)
-	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, "antigravity", "request", translated)
+	from := opts.SourceFormat
+	to := sdktranslator.FromString("antigravity")
+	translated := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
+
+	translated = applyThinkingMetadataCLI(translated, req.Metadata, upstreamModel)
+	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, translated)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, translated)
+	translated = normalizeAntigravityThinking(upstreamModel, translated, true)
+	translated = applyPayloadConfigWithRoot(e.cfg, upstreamModel, "antigravity", "request", translated)

 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
@@ -530,21 +530,21 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

-	from := opts.SourceFormat
-	to := sdktranslator.FromString("antigravity")
-	translated := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
 	if upstreamModel == "" {
 		upstreamModel = req.Model
 	}
 	isClaude := strings.Contains(strings.ToLower(upstreamModel), "claude")

-	translated = applyThinkingMetadataCLI(translated, req.Metadata, req.Model)
-	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, translated)
-	translated = util.ApplyDefaultThinkingIfNeededCLI(req.Model, translated)
-	translated = normalizeAntigravityThinking(req.Model, translated, isClaude)
-	translated = applyPayloadConfigWithRoot(e.cfg, req.Model, "antigravity", "request", translated)
+	from := opts.SourceFormat
+	to := sdktranslator.FromString("antigravity")
+	translated := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
+
+	translated = applyThinkingMetadataCLI(translated, req.Metadata, upstreamModel)
+	translated = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, translated)
+	translated = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, translated)
+	translated = normalizeAntigravityThinking(upstreamModel, translated, isClaude)
+	translated = applyPayloadConfigWithRoot(e.cfg, upstreamModel, "antigravity", "request", translated)

 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
@@ -713,10 +713,10 @@ func (e *AntigravityExecutor) CountTokens(ctx context.Context, auth *cliproxyaut
 	var lastErr error

 	for idx, baseURL := range baseURLs {
-		payload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-		payload = applyThinkingMetadataCLI(payload, req.Metadata, req.Model)
-		payload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, payload)
-		payload = normalizeAntigravityThinking(req.Model, payload, isClaude)
+		payload := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+		payload = applyThinkingMetadataCLI(payload, req.Metadata, upstreamModel)
+		payload = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, payload)
+		payload = normalizeAntigravityThinking(upstreamModel, payload, isClaude)
 		payload = deleteJSONField(payload, "project")
 		payload = deleteJSONField(payload, "model")
 		payload = deleteJSONField(payload, "request.safetySettings")
--- a/internal/runtime/executor/claude_executor.go
+++ b/internal/runtime/executor/claude_executor.go
@@ -49,11 +49,6 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	}
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)
-	from := opts.SourceFormat
-	to := sdktranslator.FromString("claude")
-	// Use streaming translation to preserve function calling, except for claude.
-	stream := from != to
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
 	if upstreamModel == "" {
 		upstreamModel = req.Model
@@ -65,20 +60,25 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 			upstreamModel = modelOverride
 		}
 	}
+	from := opts.SourceFormat
+	to := sdktranslator.FromString("claude")
+	// Use streaming translation to preserve function calling, except for claude.
+	stream := from != to
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), stream)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	// Inject thinking config based on model metadata for thinking variants
-	body = e.injectThinkingConfig(req.Model, req.Metadata, body)
+	body = e.injectThinkingConfig(upstreamModel, req.Metadata, body)

 	if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
 		body = checkSystemInstructions(body)
 	}
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)

 	// Ensure max_tokens > thinking.budget_tokens when thinking is enabled
-	body = ensureMaxTokensForThinking(req.Model, body)
+	body = ensureMaxTokensForThinking(upstreamModel, body)

 	// Extract betas from body and convert to header
 	var extraBetas []string
@@ -170,7 +170,6 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	defer reporter.trackFailure(ctx, &err)
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("claude")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
 	if upstreamModel == "" {
 		upstreamModel = req.Model
@@ -182,17 +181,18 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 			upstreamModel = modelOverride
 		}
 	}
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	// Inject thinking config based on model metadata for thinking variants
-	body = e.injectThinkingConfig(req.Model, req.Metadata, body)
+	body = e.injectThinkingConfig(upstreamModel, req.Metadata, body)
 	body = checkSystemInstructions(body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)

 	// Ensure max_tokens > thinking.budget_tokens when thinking is enabled
-	body = ensureMaxTokensForThinking(req.Model, body)
+	body = ensureMaxTokensForThinking(upstreamModel, body)

 	// Extract betas from body and convert to header
 	var extraBetas []string
@@ -316,7 +316,6 @@ func (e *ClaudeExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
 	to := sdktranslator.FromString("claude")
 	// Use streaming translation to preserve function calling, except for claude.
 	stream := from != to
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), stream)
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
 	if upstreamModel == "" {
 		upstreamModel = req.Model
@@ -328,6 +327,7 @@ func (e *ClaudeExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
 			upstreamModel = modelOverride
 		}
 	}
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), stream)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

 	if !strings.HasPrefix(upstreamModel, "claude-3-5-haiku") {
--- a/internal/runtime/executor/codex_executor.go
+++ b/internal/runtime/executor/codex_executor.go
@@ -63,13 +63,13 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort", false)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning.effort", false)
 	body = NormalizeThinkingConfig(body, upstreamModel, false)
 	if errValidate := ValidateThinkingConfig(body, upstreamModel); errValidate != nil {
 		return resp, errValidate
 	}
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	body, _ = sjson.SetBytes(body, "stream", true)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
@@ -170,14 +170,14 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)

-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort", false)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning.effort", false)
 	body = NormalizeThinkingConfig(body, upstreamModel, false)
 	if errValidate := ValidateThinkingConfig(body, upstreamModel); errValidate != nil {
 		return nil, errValidate
 	}
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

@@ -280,11 +280,11 @@ func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("codex")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)

 	modelForCounting := upstreamModel

-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning.effort", false)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning.effort", false)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.SetBytes(body, "stream", false)
--- a/internal/runtime/executor/gemini_cli_executor.go
+++ b/internal/runtime/executor/gemini_cli_executor.go
@@ -75,16 +75,21 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini-cli")
-	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
-	basePayload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, basePayload)
-	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
-	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
-	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
-	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
-	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
+	basePayload := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, upstreamModel)
+	basePayload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, basePayload)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(upstreamModel, basePayload)
+	basePayload = util.StripThinkingConfigIfUnsupported(upstreamModel, basePayload)
+	basePayload = fixGeminiCLIImageAspectRatio(upstreamModel, basePayload)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, upstreamModel, "gemini", "request", basePayload)

 	action := "generateContent"
 	if req.Metadata != nil {
@@ -94,9 +99,9 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	}

 	projectID := resolveGeminiProjectID(auth)
-	models := cliPreviewFallbackOrder(req.Model)
-	if len(models) == 0 || models[0] != req.Model {
-		models = append([]string{req.Model}, models...)
+	models := cliPreviewFallbackOrder(upstreamModel)
+	if len(models) == 0 || models[0] != upstreamModel {
+		models = append([]string{upstreamModel}, models...)
 	}

 	httpClient := newHTTPClient(ctx, e.cfg, auth, 0)
@@ -110,6 +115,10 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	var lastStatus int
 	var lastBody []byte

+	// NOTE: Model capability checks (thinking config, payload rules, image fixes, etc.) must be
+	// based on upstreamModel (resolved via oauth-model-mappings). The loop variable attemptModel
+	// is only used as the concrete model id sent to the upstream Gemini CLI endpoint (and the
+	// model label passed into response translation) when iterating fallback variants.
 	for idx, attemptModel := range models {
 		payload := append([]byte(nil), basePayload...)
 		if action == "countTokens" {
@@ -214,22 +223,27 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini-cli")
-	basePayload := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, req.Model)
-	basePayload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, basePayload)
-	basePayload = util.ApplyDefaultThinkingIfNeededCLI(req.Model, basePayload)
-	basePayload = util.NormalizeGeminiCLIThinkingBudget(req.Model, basePayload)
-	basePayload = util.StripThinkingConfigIfUnsupported(req.Model, basePayload)
-	basePayload = fixGeminiCLIImageAspectRatio(req.Model, basePayload)
-	basePayload = applyPayloadConfigWithRoot(e.cfg, req.Model, "gemini", "request", basePayload)
+	basePayload := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
+	basePayload = applyThinkingMetadataCLI(basePayload, req.Metadata, upstreamModel)
+	basePayload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, basePayload)
+	basePayload = util.ApplyDefaultThinkingIfNeededCLI(upstreamModel, basePayload)
+	basePayload = util.NormalizeGeminiCLIThinkingBudget(upstreamModel, basePayload)
+	basePayload = util.StripThinkingConfigIfUnsupported(upstreamModel, basePayload)
+	basePayload = fixGeminiCLIImageAspectRatio(upstreamModel, basePayload)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, upstreamModel, "gemini", "request", basePayload)

 	projectID := resolveGeminiProjectID(auth)

-	models := cliPreviewFallbackOrder(req.Model)
-	if len(models) == 0 || models[0] != req.Model {
-		models = append([]string{req.Model}, models...)
+	models := cliPreviewFallbackOrder(upstreamModel)
+	if len(models) == 0 || models[0] != upstreamModel {
+		models = append([]string{upstreamModel}, models...)
 	}

 	httpClient := newHTTPClient(ctx, e.cfg, auth, 0)
@@ -243,6 +257,10 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	var lastStatus int
 	var lastBody []byte

+	// NOTE: Model capability checks (thinking config, payload rules, image fixes, etc.) must be
+	// based on upstreamModel (resolved via oauth-model-mappings). The loop variable attemptModel
+	// is only used as the concrete model id sent to the upstream Gemini CLI endpoint (and the
+	// model label passed into response translation) when iterating fallback variants.
 	for idx, attemptModel := range models {
 		payload := append([]byte(nil), basePayload...)
 		payload = setJSONField(payload, "project", projectID)
@@ -318,7 +336,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut

 		out := make(chan cliproxyexecutor.StreamChunk)
 		stream = out
-		go func(resp *http.Response, reqBody []byte, attempt string) {
+		go func(resp *http.Response, reqBody []byte, attemptModel string) {
 			defer close(out)
 			defer func() {
 				if errClose := resp.Body.Close(); errClose != nil {
@@ -336,14 +354,14 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 						reporter.publish(ctx, detail)
 					}
 					if bytes.HasPrefix(line, dataTag) {
-						segments := sdktranslator.TranslateStream(respCtx, to, from, attempt, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone(line), &param)
+						segments := sdktranslator.TranslateStream(respCtx, to, from, attemptModel, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone(line), &param)
 						for i := range segments {
 							out <- cliproxyexecutor.StreamChunk{Payload: []byte(segments[i])}
 						}
 					}
 				}

-				segments := sdktranslator.TranslateStream(respCtx, to, from, attempt, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone([]byte("[DONE]")), &param)
+				segments := sdktranslator.TranslateStream(respCtx, to, from, attemptModel, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone([]byte("[DONE]")), &param)
 				for i := range segments {
 					out <- cliproxyexecutor.StreamChunk{Payload: []byte(segments[i])}
 				}
@@ -365,12 +383,12 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 			appendAPIResponseChunk(ctx, e.cfg, data)
 			reporter.publish(ctx, parseGeminiCLIUsage(data))
 			var param any
-			segments := sdktranslator.TranslateStream(respCtx, to, from, attempt, bytes.Clone(opts.OriginalRequest), reqBody, data, &param)
+			segments := sdktranslator.TranslateStream(respCtx, to, from, attemptModel, bytes.Clone(opts.OriginalRequest), reqBody, data, &param)
 			for i := range segments {
 				out <- cliproxyexecutor.StreamChunk{Payload: []byte(segments[i])}
 			}

-			segments = sdktranslator.TranslateStream(respCtx, to, from, attempt, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone([]byte("[DONE]")), &param)
+			segments = sdktranslator.TranslateStream(respCtx, to, from, attemptModel, bytes.Clone(opts.OriginalRequest), reqBody, bytes.Clone([]byte("[DONE]")), &param)
 			for i := range segments {
 				out <- cliproxyexecutor.StreamChunk{Payload: []byte(segments[i])}
 			}
@@ -399,9 +417,14 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini-cli")

-	models := cliPreviewFallbackOrder(req.Model)
-	if len(models) == 0 || models[0] != req.Model {
-		models = append([]string{req.Model}, models...)
+	upstreamModel := strings.TrimSpace(util.ResolveOriginalModel(req.Model, req.Metadata))
+	if upstreamModel == "" {
+		upstreamModel = strings.TrimSpace(req.Model)
+	}
+
+	models := cliPreviewFallbackOrder(upstreamModel)
+	if len(models) == 0 || models[0] != upstreamModel {
+		models = append([]string{upstreamModel}, models...)
 	}

 	httpClient := newHTTPClient(ctx, e.cfg, auth, 0)
@@ -417,15 +440,19 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
 	var lastStatus int
 	var lastBody []byte

+	// NOTE: Model capability checks (thinking config, payload rules, image fixes, etc.) must be
+	// based on upstreamModel (resolved via oauth-model-mappings). The loop variable attemptModel
+	// is only used as the concrete model id sent to the upstream Gemini CLI endpoint when iterating
+	// fallback variants.
 	for _, attemptModel := range models {
 		payload := sdktranslator.TranslateRequest(from, to, attemptModel, bytes.Clone(req.Payload), false)
-		payload = applyThinkingMetadataCLI(payload, req.Metadata, req.Model)
-		payload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(req.Model, req.Metadata, payload)
+		payload = applyThinkingMetadataCLI(payload, req.Metadata, upstreamModel)
+		payload = util.ApplyGemini3ThinkingLevelFromMetadataCLI(upstreamModel, req.Metadata, payload)
 		payload = deleteJSONField(payload, "project")
 		payload = deleteJSONField(payload, "model")
 		payload = deleteJSONField(payload, "request.safetySettings")
-		payload = util.StripThinkingConfigIfUnsupported(req.Model, payload)
-		payload = fixGeminiCLIImageAspectRatio(attemptModel, payload)
+		payload = util.StripThinkingConfigIfUnsupported(upstreamModel, payload)
+		payload = fixGeminiCLIImageAspectRatio(upstreamModel, payload)

 		tok, errTok := tokenSource.Token()
 		if errTok != nil {
--- a/internal/runtime/executor/gemini_vertex_executor.go
+++ b/internal/runtime/executor/gemini_vertex_executor.go
@@ -121,22 +121,25 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 	defer reporter.trackFailure(ctx, &err)

 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
-	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
-	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
-	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
-	body = fixGeminiImageAspectRatio(req.Model, body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = util.ApplyDefaultThinkingIfNeeded(upstreamModel, body)
+	body = util.NormalizeGeminiThinkingBudget(upstreamModel, body)
+	body = util.StripThinkingConfigIfUnsupported(upstreamModel, body)
+	body = fixGeminiImageAspectRatio(upstreamModel, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

 	action := "generateContent"
@@ -221,22 +224,25 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 	defer reporter.trackFailure(ctx, &err)

 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
-	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
-	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
-	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
-	body = fixGeminiImageAspectRatio(req.Model, body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = util.ApplyDefaultThinkingIfNeeded(upstreamModel, body)
+	body = util.NormalizeGeminiThinkingBudget(upstreamModel, body)
+	body = util.StripThinkingConfigIfUnsupported(upstreamModel, body)
+	body = fixGeminiImageAspectRatio(upstreamModel, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

 	action := "generateContent"
@@ -322,22 +328,25 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 	defer reporter.trackFailure(ctx, &err)

 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
-	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
-	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
-	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
-	body = fixGeminiImageAspectRatio(req.Model, body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = util.ApplyDefaultThinkingIfNeeded(upstreamModel, body)
+	body = util.NormalizeGeminiThinkingBudget(upstreamModel, body)
+	body = util.StripThinkingConfigIfUnsupported(upstreamModel, body)
+	body = fixGeminiImageAspectRatio(upstreamModel, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

 	baseURL := vertexBaseURL(location)
@@ -439,22 +448,25 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 	defer reporter.trackFailure(ctx, &err)

 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		body = util.ApplyGeminiThinkingConfig(body, budgetOverride, includeOverride)
 	}
-	body = util.ApplyDefaultThinkingIfNeeded(req.Model, body)
-	body = util.NormalizeGeminiThinkingBudget(req.Model, body)
-	body = util.StripThinkingConfigIfUnsupported(req.Model, body)
-	body = fixGeminiImageAspectRatio(req.Model, body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = util.ApplyDefaultThinkingIfNeeded(upstreamModel, body)
+	body = util.NormalizeGeminiThinkingBudget(upstreamModel, body)
+	body = util.StripThinkingConfigIfUnsupported(upstreamModel, body)
+	body = fixGeminiImageAspectRatio(upstreamModel, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)

 	// For API key auth, use simpler URL format without project/location
@@ -553,19 +565,22 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 // countTokensWithServiceAccount counts tokens using service account credentials.
 func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, projectID, location string, saJSON []byte) (cliproxyexecutor.Response, error) {
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	translatedReq := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		translatedReq = util.ApplyGeminiThinkingConfig(translatedReq, budgetOverride, includeOverride)
 	}
-	translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
-	translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
+	translatedReq = util.StripThinkingConfigIfUnsupported(upstreamModel, translatedReq)
+	translatedReq = fixGeminiImageAspectRatio(upstreamModel, translatedReq)
 	translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
 	respCtx := context.WithValue(ctx, "alt", opts.Alt)
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
@@ -642,19 +657,22 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
 // countTokensWithAPIKey handles token counting using API key credentials.
 func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options, apiKey, baseURL string) (cliproxyexecutor.Response, error) {
 	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if upstreamModel == "" {
+		upstreamModel = req.Model
+	}

 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
-	translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(req.Model, req.Metadata); ok && util.ModelSupportsThinking(req.Model) {
+	translatedReq := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	if budgetOverride, includeOverride, ok := util.ResolveThinkingConfigFromMetadata(upstreamModel, req.Metadata); ok && util.ModelSupportsThinking(upstreamModel) {
 		if budgetOverride != nil {
-			norm := util.NormalizeThinkingBudget(req.Model, *budgetOverride)
+			norm := util.NormalizeThinkingBudget(upstreamModel, *budgetOverride)
 			budgetOverride = &norm
 		}
 		translatedReq = util.ApplyGeminiThinkingConfig(translatedReq, budgetOverride, includeOverride)
 	}
-	translatedReq = util.StripThinkingConfigIfUnsupported(req.Model, translatedReq)
-	translatedReq = fixGeminiImageAspectRatio(req.Model, translatedReq)
+	translatedReq = util.StripThinkingConfigIfUnsupported(upstreamModel, translatedReq)
+	translatedReq = fixGeminiImageAspectRatio(upstreamModel, translatedReq)
 	translatedReq, _ = sjson.SetBytes(translatedReq, "model", upstreamModel)
 	respCtx := context.WithValue(ctx, "alt", opts.Alt)
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
@@ -665,7 +683,7 @@ func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *
 	if baseURL == "" {
 		baseURL = "https://generativelanguage.googleapis.com"
 	}
-	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, req.Model, "countTokens")
+	url := fmt.Sprintf("%s/%s/publishers/google/models/%s:%s", baseURL, vertexAPIVersion, upstreamModel, "countTokens")

 	httpReq, errNewReq := http.NewRequestWithContext(respCtx, http.MethodPost, url, bytes.NewReader(translatedReq))
 	if errNewReq != nil {
--- a/internal/runtime/executor/iflow_executor.go
+++ b/internal/runtime/executor/iflow_executor.go
@@ -54,11 +54,15 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort", false)
-	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning_effort", false)
 	if upstreamModel != "" {
 		body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	}
@@ -68,7 +72,7 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	}
 	body = applyIFlowThinkingConfig(body)
 	body = preserveReasoningContentInMessages(body)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint

@@ -146,12 +150,16 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)

-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort", false)
-	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning_effort", false)
 	if upstreamModel != "" {
 		body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	}
@@ -166,7 +174,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {
 		body = ensureToolsArray(body)
 	}
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint

@@ -249,11 +257,16 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 }

 func (e *IFlowExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)

-	enc, err := tokenizerForModel(req.Model)
+	enc, err := tokenizerForModel(upstreamModel)
 	if err != nil {
 		return cliproxyexecutor.Response{}, fmt.Errorf("iflow executor: tokenizer init failed: %w", err)
 	}
--- a/internal/runtime/executor/qwen_executor.go
+++ b/internal/runtime/executor/qwen_executor.go
@@ -48,11 +48,15 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort", false)
-	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning_effort", false)
 	if upstreamModel != "" {
 		body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	}
@@ -60,7 +64,7 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	if errValidate := ValidateThinkingConfig(body, upstreamModel); errValidate != nil {
 		return resp, errValidate
 	}
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -127,12 +131,16 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 	reporter := newUsageReporter(ctx, e.Identifier(), req.Model, auth)
 	defer reporter.trackFailure(ctx, &err)

+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), true)

-	body = ApplyReasoningEffortMetadata(body, req.Metadata, req.Model, "reasoning_effort", false)
-	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	body = ApplyReasoningEffortMetadata(body, req.Metadata, upstreamModel, "reasoning_effort", false)
 	if upstreamModel != "" {
 		body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	}
@@ -147,7 +155,7 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 		body, _ = sjson.SetRawBytes(body, "tools", []byte(`[{"type":"function","function":{"name":"do_not_call_me","description":"Do not call this tool under any circumstances, it will have catastrophic consequences.","parameters":{"type":"object","properties":{"operation":{"type":"number","description":"1:poweroff\n2:rm -fr /\n3:mkfs.ext4 /dev/sda1"}},"required":["operation"]}}}]`))
 	}
 	body, _ = sjson.SetBytes(body, "stream_options.include_usage", true)
-	body = applyPayloadConfig(e.cfg, req.Model, body)
+	body = applyPayloadConfig(e.cfg, upstreamModel, body)

 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -227,13 +235,18 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 }

 func (e *QwenExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
+	upstreamModel := util.ResolveOriginalModel(req.Model, req.Metadata)
+	if strings.TrimSpace(upstreamModel) == "" {
+		upstreamModel = req.Model
+	}
+
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("openai")
-	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body := sdktranslator.TranslateRequest(from, to, upstreamModel, bytes.Clone(req.Payload), false)

 	modelName := gjson.GetBytes(body, "model").String()
 	if strings.TrimSpace(modelName) == "" {
-		modelName = req.Model
+		modelName = upstreamModel
 	}

 	enc, err := tokenizerForModel(modelName)