gcc-16/changes.html - GCN: offload launch overhead reduced

diff --git a/htdocs/gcc-16/changes.html b/htdocs/gcc-16/changes.html
index a921e805..500ab53a 100644
--- a/htdocs/gcc-16/changes.html
+++ b/htdocs/gcc-16/changes.html
@@ -436,6 +436,8 @@ struct Affected : E
   <li>Experimental support for AMD Instinct MI300 (<code>gfx942</code>) devices
       has been added, including the generic <code>gfx9-4-generic</code> and
       mostly compatible <code>gfx950</code>.</li>
+  <li>For offloading to AMD GPUs: The launch overhead of OpenMP target regions
+      and OpenACC compute regions has been drastically reduced.</li>
   <li>By default, the following multilibs are now built: <code>gfx908</code>,
       <code>gfx90a</code>, <code>gfx9-generic</code>,
       <code>gfx9-4-generic</code>, <code>gfx10-3-generic</code>, and