diff --git a/codex-rs/core/src/codex/compact.rs b/codex-rs/core/src/codex/compact.rs
index b6bca43e10..d1e12cc6bc 100644
--- a/codex-rs/core/src/codex/compact.rs
+++ b/codex-rs/core/src/codex/compact.rs
@@ -78,7 +78,8 @@ async fn run_compact_task_inner(
     let mut trimmed_tails: Vec<Vec<ResponseItem>> = Vec::new();
 
     let max_retries = turn_context.client.get_provider().stream_max_retries();
-    let mut retries = 0;
+    let mut context_retries = 0u64;
+    let mut stream_retries = 0u64;
 
     let rollout_item = RolloutItem::TurnContext(TurnContextItem {
         cwd: turn_context.cwd.clone(),
@@ -124,8 +125,7 @@ async fn run_compact_task_inner(
                     if !trimmed.is_empty() {
                         truncated_count += trimmed.len();
                         trimmed_tails.push(trimmed);
-                        retries += 1;
-                        if retries >= max_retries {
+                        if context_retries >= max_retries {
                             sess.set_total_tokens_full(&sub_id, turn_context.as_ref())
                                 .await;
                             let event = Event {
@@ -137,6 +137,9 @@ async fn run_compact_task_inner(
                             sess.send_event(event).await;
                             return;
                         }
+                        context_retries += 1;
+                        stream_retries = 0;
+                        // Keep stream retry budget untouched; we trimmed context successfully.
                         continue;
                     }
                 }
@@ -152,12 +155,12 @@ async fn run_compact_task_inner(
                 return;
             }
             Err(e) => {
-                if retries < max_retries {
-                    retries += 1;
-                    let delay = backoff(retries);
+                if stream_retries < max_retries {
+                    stream_retries += 1;
+                    let delay = backoff(stream_retries);
                     sess.notify_stream_error(
                         &sub_id,
-                        format!("Re-connecting... {retries}/{max_retries}"),
+                        format!("Re-connecting... {stream_retries}/{max_retries}"),
                     )
                     .await;
                     tokio::time::sleep(delay).await;
diff --git a/codex-rs/core/tests/suite/compact.rs b/codex-rs/core/tests/suite/compact.rs
index 518888f81a..d1e0bb2d6a 100644
--- a/codex-rs/core/tests/suite/compact.rs
+++ b/codex-rs/core/tests/suite/compact.rs
@@ -32,6 +32,7 @@ use serde_json::Value;
 pub(super) const FIRST_REPLY: &str = "FIRST_REPLY";
 pub(super) const SUMMARY_TEXT: &str = "SUMMARY_ONLY_CONTEXT";
 const THIRD_USER_MSG: &str = "next turn";
+const THIRD_ASSISTANT_MSG: &str = "post compact assistant";
 const AUTO_SUMMARY_TEXT: &str = "AUTO_SUMMARY";
 const FIRST_AUTO_MSG: &str = "token limit start";
 const SECOND_AUTO_MSG: &str = "token limit push";
@@ -646,6 +647,10 @@ async fn manual_compact_retries_after_context_window_error() {
         ev_assistant_message("m2", SUMMARY_TEXT),
         ev_completed("r2"),
     ]);
+    let third_turn = sse(vec![
+        ev_assistant_message("m3", THIRD_ASSISTANT_MSG),
+        ev_completed("r3"),
+    ]);
 
     let request_log = mount_sse_sequence(
         &server,
@@ -653,6 +658,7 @@ async fn manual_compact_retries_after_context_window_error() {
             user_turn.clone(),
             compact_failed.clone(),
             compact_succeeds.clone(),
+            third_turn,
         ],
     )
     .await;
@@ -711,8 +717,8 @@ async fn manual_compact_retries_after_context_window_error() {
     let requests = request_log.requests();
     assert_eq!(
         requests.len(),
-        3,
-        "expected user turn and two compact attempts"
+        4,
+        "expected user turn, two compact attempts, and one follow-up turn"
     );
 
     let compact_attempt = requests[1].body_json();