diff --git a/core/http/middleware/request.go b/core/http/middleware/request.go
index e60aa33939..8bddd002a5 100644
--- a/core/http/middleware/request.go
+++ b/core/http/middleware/request.go
@@ -862,6 +862,17 @@ func MergeOpenResponsesConfig(config *config.ModelConfig, input *schema.OpenResp
 		config.Maxtokens = input.MaxOutputTokens
 	}
 
+	// reasoning.effort and request metadata reach the backend the same way as
+	// on chat/completions (see mergeOpenAIRequestAndModelConfig).
+	var effort string
+	if input.Reasoning != nil {
+		effort = input.Reasoning.Effort
+	}
+	config.ApplyReasoningEffort(effort)
+	if len(input.Metadata) > 0 {
+		config.RequestMetadata = input.Metadata
+	}
+
 	// Convert tools to functions - this will be handled in the endpoint handler
 	// We just validate that tools are present if needed
 
diff --git a/core/http/middleware/request_test.go b/core/http/middleware/request_test.go
index afaf8d9c80..d2d1d539df 100644
--- a/core/http/middleware/request_test.go
+++ b/core/http/middleware/request_test.go
@@ -426,6 +426,21 @@ var _ = Describe("MergeOpenResponsesConfig tool_choice parsing", func() {
 	})
 })
 
+var _ = Describe("MergeOpenResponsesConfig reasoning", func() {
+	It("applies reasoning.effort and forwards request metadata", func() {
+		cfg := &config.ModelConfig{}
+		req := &schema.OpenResponsesRequest{
+			Reasoning: &schema.ORReasoningParam{Effort: "none"},
+			Metadata:  map[string]string{"enable_thinking": "false"},
+		}
+		Expect(MergeOpenResponsesConfig(cfg, req)).To(Succeed())
+
+		Expect(cfg.ReasoningEffort).To(Equal("none"))
+		Expect(cfg.ReasoningConfig.DisableReasoning).To(HaveValue(BeTrue()))
+		Expect(cfg.RequestMetadata).To(Equal(req.Metadata))
+	})
+})
+
 // ---------------------------------------------------------------------------
 // SetModelAndConfig + SetOpenAIRequest - /v1/chat/completions tool_choice parsing
 // ---------------------------------------------------------------------------