diff --git a/internal/js/chat-stream/sse_parse_impl.js b/internal/js/chat-stream/sse_parse_impl.js
index 09530bc..834c392 100644
--- a/internal/js/chat-stream/sse_parse_impl.js
+++ b/internal/js/chat-stream/sse_parse_impl.js
@@ -59,7 +59,7 @@ function parseChunkForContent(chunk, thinkingEnabled, currentType, stripReferenc
     };
   }
   if (isStatusPath(pathValue)) {
-    if (asString(chunk.v) === 'FINISHED') {
+    if (isFinishedStatus(chunk.v)) {
       return {
         parsed: true,
         parts: [],
@@ -149,7 +149,7 @@ function parseChunkForContent(chunk, thinkingEnabled, currentType, stripReferenc
 
   const val = chunk.v;
   if (typeof val === 'string') {
-    if (val === 'FINISHED' && (!pathValue || pathValue === 'status')) {
+    if (isFinishedStatus(val) && (!pathValue || pathValue === 'status')) {
       return {
         parsed: true,
         parts: [],
@@ -258,7 +258,7 @@ function extractContentRecursive(items, defaultType, stripReferenceMarkers = tru
     const itemPath = asString(it.p);
     const itemV = it.v;
     if (isStatusPath(itemPath)) {
-      if (asString(itemV) === 'FINISHED') {
+      if (isFinishedStatus(itemV)) {
         return { parts: [], finished: true };
       }
       continue;
@@ -336,6 +336,10 @@ function isStatusPath(pathValue) {
   return pathValue === 'response/status' || pathValue === 'status';
 }
 
+function isFinishedStatus(value) {
+  return asString(value).toUpperCase() === 'FINISHED';
+}
+
 function filterLeakedContentFilterParts(parts) {
   if (!Array.isArray(parts) || parts.length === 0) {
     return parts;
diff --git a/tests/node/chat-stream.test.js b/tests/node/chat-stream.test.js
index 6086ba1..205cdeb 100644
--- a/tests/node/chat-stream.test.js
+++ b/tests/node/chat-stream.test.js
@@ -291,6 +291,19 @@ test('parseChunkForContent preserves output tokens on FINISHED lines', () => {
   assert.deepEqual(parsed.parts, []);
 });
 
+test('parseChunkForContent matches FINISHED case-insensitively on status paths', () => {
+  const parsed = parseChunkForContent(
+    { p: 'response/status', v: ' finished ', accumulated_token_usage: 190 },
+    false,
+    'text',
+  );
+  assert.equal(parsed.parsed, true);
+  assert.equal(parsed.finished, true);
+  assert.equal(parsed.contentFilter, false);
+  assert.equal(parsed.outputTokens, 190);
+  assert.deepEqual(parsed.parts, []);
+});
+
 test('parseChunkForContent filters INCOMPLETE status text without stopping stream', () => {
   const parsed = parseChunkForContent(
     { p: 'response/status', v: 'INCOMPLETE', accumulated_token_usage: 190 },