diff --git a/src/providers/gemini.rs b/src/providers/gemini.rs
index b3b7110a9..49d10696a 100644
--- a/src/providers/gemini.rs
+++ b/src/providers/gemini.rs
@@ -72,6 +72,8 @@ struct GenerateContentRequest {
 #[derive(Debug, Serialize)]
 struct InternalGenerateContentEnvelope {
     model: String,
+    generation_config: InternalGenerationConfig,
+    contents: Vec<Content>,
     #[serde(skip_serializing_if = "Option::is_none")]
     project: Option<String>,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -108,6 +110,12 @@ struct GenerationConfig {
     max_output_tokens: u32,
 }
 
+#[derive(Debug, Serialize, Clone)]
+struct InternalGenerationConfig {
+    temperature: f64,
+    max_output_tokens: u32,
+}
+
 #[derive(Debug, Deserialize)]
 struct GenerateContentResponse {
     candidates: Option<Vec<Candidate>>,
@@ -316,16 +324,37 @@ impl GeminiProvider {
         let req = self.http_client().post(url).json(request);
         match auth {
             GeminiAuth::OAuthToken(token) => {
-                // cloudcode-pa expects an outer envelope with `request`.
-                let internal_request = InternalGenerateContentEnvelope {
-                    model: Self::format_internal_model_name(model),
-                    project: None,
-                    user_prompt_id: None,
-                    request: InternalGenerateContentRequest {
-                        contents: request.contents.clone(),
-                        system_instruction: request.system_instruction.clone(),
-                        generation_config: request.generation_config.clone(),
+                // Internal API expects the model in the request body envelope
+                let internal_request = InternalGenerateContentRequest {
+                    model: Self::format_model_name(model),
+                    generation_config: InternalGenerationConfig {
+                        temperature: request.generation_config.temperature,
+                        max_output_tokens: request.generation_config.max_output_tokens,
                     },
+                    contents: request
+                        .contents
+                        .iter()
+                        .map(|c| Content {
+                            role: c.role.clone(),
+                            parts: c
+                                .parts
+                                .iter()
+                                .map(|p| Part {
+                                    text: p.text.clone(),
+                                })
+                                .collect(),
+                        })
+                        .collect(),
+                    system_instruction: request.system_instruction.as_ref().map(|si| Content {
+                        role: si.role.clone(),
+                        parts: si
+                            .parts
+                            .iter()
+                            .map(|p| Part {
+                                text: p.text.clone(),
+                            })
+                            .collect(),
+                    }),
                 };
                 self.http_client()
                     .post(url)
@@ -747,16 +776,16 @@ mod tests {
 
     #[test]
     fn internal_request_includes_model() {
-        let request = InternalGenerateContentEnvelope {
-            model: "gemini-test-model".to_string(),
-            project: None,
-            user_prompt_id: None,
-            request: InternalGenerateContentRequest {
-                contents: vec![Content {
-                    role: Some("user".to_string()),
-                    parts: vec![Part {
-                        text: "Hello".to_string(),
-                    }],
+        let request = InternalGenerateContentRequest {
+            model: "models/gemini-3-pro-preview".to_string(),
+            generation_config: InternalGenerationConfig {
+                temperature: 0.7,
+                max_output_tokens: 8192,
+            },
+            contents: vec![Content {
+                role: Some("user".to_string()),
+                parts: vec![Part {
+                    text: "Hello".to_string(),
                 }],
                 system_instruction: None,
                 generation_config: GenerationConfig {
@@ -766,11 +795,12 @@ mod tests {
             },
         };
 
-        let json: serde_json::Value = serde_json::to_value(&request).unwrap();
-        assert_eq!(json["model"], "gemini-test-model");
-        assert!(json.get("generationConfig").is_none());
-        assert!(json["request"].get("generationConfig").is_some());
-        assert_eq!(json["request"]["contents"][0]["role"], "user");
+        let json = serde_json::to_string(&request).unwrap();
+        assert!(json.contains("\"model\":\"models/gemini-3-pro-preview\""));
+        assert!(json.contains("\"generation_config\""));
+        assert!(json.contains("\"max_output_tokens\":8192"));
+        assert!(json.contains("\"role\":\"user\""));
+        assert!(json.contains("\"temperature\":0.7"));
     }
 
     #[test]