diff --git a/internal/runtime/executor/codex_executor.go b/internal/runtime/executor/codex_executor.go
index 3fe5ed6e..7003373f 100644
--- a/internal/runtime/executor/codex_executor.go
+++ b/internal/runtime/executor/codex_executor.go
@@ -56,6 +56,9 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
 	body = applyReasoningEffortMetadata(body, req.Metadata, req.Model)
 	body = normalizeThinkingConfig(body, upstreamModel)
+	if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
+		return resp, errValidate
+	}
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
 	body, _ = sjson.SetBytes(body, "stream", true)
@@ -151,6 +154,9 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 
 	body = applyReasoningEffortMetadata(body, req.Metadata, req.Model)
 	body = normalizeThinkingConfig(body, upstreamModel)
+	if errValidate := validateThinkingConfig(body, upstreamModel); errValidate != nil {
+		return nil, errValidate
+	}
 	body = applyPayloadConfig(e.cfg, req.Model, body)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.SetBytes(body, "model", upstreamModel)
diff --git a/internal/runtime/executor/openai_compat_executor.go b/internal/runtime/executor/openai_compat_executor.go
index ba47750e..507b0fd9 100644
--- a/internal/runtime/executor/openai_compat_executor.go
+++ b/internal/runtime/executor/openai_compat_executor.go
@@ -64,6 +64,9 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 		translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
 	}
 	translated = normalizeThinkingConfig(translated, upstreamModel)
+	if errValidate := validateThinkingConfig(translated, upstreamModel); errValidate != nil {
+		return resp, errValidate
+	}
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
@@ -155,6 +158,9 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 		translated, _ = sjson.SetBytes(translated, "model", upstreamModel)
 	}
 	translated = normalizeThinkingConfig(translated, upstreamModel)
+	if errValidate := validateThinkingConfig(translated, upstreamModel); errValidate != nil {
+		return nil, errValidate
+	}
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(translated))
diff --git a/internal/runtime/executor/payload_helpers.go b/internal/runtime/executor/payload_helpers.go
index 9d431f11..5711bbbd 100644
--- a/internal/runtime/executor/payload_helpers.go
+++ b/internal/runtime/executor/payload_helpers.go
@@ -1,6 +1,8 @@
 package executor
 
 import (
+	"fmt"
+	"net/http"
 	"strings"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
@@ -289,3 +291,36 @@ func normalizeReasoningEffortLevel(payload []byte, model string) []byte {
 
 	return out
 }
+
+// validateThinkingConfig checks for unsupported reasoning levels on level-based models.
+// Returns a statusErr with 400 when an unsupported level is supplied to avoid silently
+// downgrading requests.
+func validateThinkingConfig(payload []byte, model string) error {
+	if len(payload) == 0 || model == "" {
+		return nil
+	}
+	if !util.ModelSupportsThinking(model) || !util.ModelUsesThinkingLevels(model) {
+		return nil
+	}
+
+	levels := util.GetModelThinkingLevels(model)
+	checkField := func(path string) error {
+		if effort := gjson.GetBytes(payload, path); effort.Exists() {
+			if _, ok := util.NormalizeReasoningEffortLevel(model, effort.String()); !ok {
+				return statusErr{
+					code: http.StatusBadRequest,
+					msg:  fmt.Sprintf("unsupported reasoning effort level %q for model %s (supported: %s)", effort.String(), model, strings.Join(levels, ", ")),
+				}
+			}
+		}
+		return nil
+	}
+
+	if err := checkField("reasoning_effort"); err != nil {
+		return err
+	}
+	if err := checkField("reasoning.effort"); err != nil {
+		return err
+	}
+	return nil
+}
diff --git a/internal/util/thinking.go b/internal/util/thinking.go
index 37200980..9671f20b 100644
--- a/internal/util/thinking.go
+++ b/internal/util/thinking.go
@@ -91,8 +91,7 @@ func ModelUsesThinkingLevels(model string) bool {
 }
 
 // NormalizeReasoningEffortLevel validates and normalizes a reasoning effort
-// level for the given model. If the level is not supported, it returns the
-// first (lowest) level from the model's supported levels.
+// level for the given model. Returns false when the level is not supported.
 func NormalizeReasoningEffortLevel(model, effort string) (string, bool) {
 	levels := GetModelThinkingLevels(model)
 	if len(levels) == 0 {
@@ -104,12 +103,5 @@ func NormalizeReasoningEffortLevel(model, effort string) (string, bool) {
 			return lvl, true
 		}
 	}
-	return defaultReasoningLevel(levels), true
-}
-
-func defaultReasoningLevel(levels []string) string {
-	if len(levels) > 0 {
-		return levels[0]
-	}
-	return ""
+	return "", false
 }