From da6d549ab28a767f73f325db4d20d7c0e3d7cafa Mon Sep 17 00:00:00 2001
From: Faradawn Yang <73060648+faradawn@users.noreply.github.com>
Date: Wed, 20 May 2026 01:44:54 -0500
Subject: [PATCH] Update GLM-5 H200 FP8 (#25814)

Signed-off-by: Faradawn Yang <73060648+faradawn@users.noreply.github.com>
---
 docs_new/cookbook/autoregressive/GLM/GLM-5.mdx            | 1 +
 docs_new/src/snippets/autoregressive/glm-5-deployment.jsx | 5 +++++
 2 files changed, 6 insertions(+)

diff --git a/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx b/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx
index 406e64aa1..b937d71a8 100644
--- a/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx
+++ b/docs_new/cookbook/autoregressive/GLM/GLM-5.mdx
@@ -109,6 +109,7 @@ sglang serve \
   --speculative-num-steps 3 \
   --speculative-eagle-topk 1 \
   --speculative-num-draft-tokens 4 \
+  --enable-flashinfer-allreduce-fusion \
   --mem-fraction-static 0.85 \
   --host 0.0.0.0 \
   --port 30000
diff --git a/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx b/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx
index 26f1931ad..e7f8c9550 100644
--- a/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx
+++ b/docs_new/src/snippets/autoregressive/glm-5-deployment.jsx
@@ -206,6 +206,11 @@ export const GLM5Deployment = () => {
       cmd += ' \\\n  --enable-flashinfer-allreduce-fusion';
     }
 
+    // H200 FP8: flashinfer allreduce fusion.
+    if (hardware === 'h200' && effectiveQuant === 'fp8') {
+      cmd += ' \\\n  --enable-flashinfer-allreduce-fusion';
+    }
+
     cmd += ` \\\n  --mem-fraction-static ${memFraction}`;
     return cmd;
   };