support 'flex' tier in app-server in addition to 'fast' (#13391)

codex-rs/core/src/client.rs

@@ -546,7 +546,8 @@ impl ModelClientSession {
             include,
             service_tier: match service_tier {
                 Some(ServiceTier::Fast) => Some("priority".to_string()),
-                _ => None,
+                Some(service_tier) => Some(service_tier.to_string()),
+                None => None,
             },
             prompt_cache_key,
             text,

codex-rs/core/src/client_common.rs

@@ -234,6 +234,7 @@ mod tests {
     use codex_api::common::OpenAiVerbosity;
     use codex_api::common::TextControls;
     use codex_api::create_text_param_for_request;
+    use codex_protocol::config_types::ServiceTier;
     use codex_protocol::models::FunctionCallOutputPayload;
     use pretty_assertions::assert_eq;
 
@@ -342,6 +343,31 @@ mod tests {
         assert!(v.get("text").is_none());
     }
 
+    #[test]
+    fn serializes_flex_service_tier_when_set() {
+        let req = ResponsesApiRequest {
+            model: "gpt-5.1".to_string(),
+            instructions: "i".to_string(),
+            input: vec![],
+            tools: vec![],
+            tool_choice: "auto".to_string(),
+            parallel_tool_calls: true,
+            reasoning: None,
+            store: false,
+            stream: true,
+            include: vec![],
+            prompt_cache_key: None,
+            service_tier: Some(ServiceTier::Flex.to_string()),
+            text: None,
+        };
+
+        let v = serde_json::to_value(&req).expect("json");
+        assert_eq!(
+            v.get("service_tier").and_then(|tier| tier.as_str()),
+            Some("flex")
+        );
+    }
+
     #[test]
     fn reserializes_shell_outputs_for_function_and_custom_tool_calls() {
         let raw_output = r#"{"output":"hello","metadata":{"exit_code":0,"duration_seconds":0.5}}"#;

codex-rs/core/src/config/mod.rs

@@ -189,7 +189,7 @@ pub struct Config {
     /// Optional override of model selection.
     pub model: Option<String>,
 
-    /// Effective service tier preference for new turns.
+    /// Effective service tier preference for new turns (`fast` or `flex`).
     pub service_tier: Option<ServiceTier>,
 
     /// Model used specifically for review sessions.
@@ -1191,7 +1191,7 @@ pub struct ConfigToml {
     /// Optionally specify a personality for the model
     pub personality: Option<Personality>,
 
-    /// Optional explicit service tier preference for new turns.
+    /// Optional explicit service tier preference for new turns (`fast` or `flex`).
     pub service_tier: Option<ServiceTier>,
 
     /// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
@@ -1984,15 +1984,14 @@ impl Config {
         let forced_login_method = cfg.forced_login_method;
 
         let model = model.or(config_profile.model).or(cfg.model);
-        let service_tier = if features.enabled(Feature::FastMode) {
-            service_tier_override.unwrap_or_else(|| {
-                config_profile
-                    .service_tier
-                    .or(cfg.service_tier)
-                    .filter(|tier| matches!(tier, ServiceTier::Fast))
-            })
-        } else {
-            None
+        let service_tier = service_tier_override
+            .unwrap_or_else(|| config_profile.service_tier.or(cfg.service_tier));
+        let service_tier = match service_tier {
+            Some(ServiceTier::Fast) if features.enabled(Feature::FastMode) => {
+                Some(ServiceTier::Fast)
+            }
+            Some(ServiceTier::Flex) => Some(ServiceTier::Flex),
+            _ => None,
         };
 
         let compact_prompt = compact_prompt.or(cfg.compact_prompt).and_then(|value| {