diff --git a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p4d-dep8-tp4-c48-mtp.yaml b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p4d-dep8-tp4-c48-mtp.yaml
index e779088aa2..f867233dbe 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p4d-dep8-tp4-c48-mtp.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p4d-dep8-tp4-c48-mtp.yaml
@@ -8,10 +8,10 @@ identity:
     repo: nvidia/GLM-5.2-NVFP4
     revision: aec724e8c7b8ee9db3b48c01c320f63f9cdaf8aa
   container:
-    image: lmsysorg/sglang:nightly-dev-cu13-20260805-211ee642
+    image: lmsysorg/sglang:v0.5.19-cu130@sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9
   frameworks:
-    dynamo: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
-    sglang: nightly-dev-cu13-20260805-211ee642
+    dynamo: "1.5.0.dev20260910"
+    sglang: "0.5.19"
 resources:
   gpu_type: gb200
   gpus_per_node: 4
@@ -35,8 +35,8 @@ frontend:
     active-prefill-tokens-threshold: None
     active-prefill-tokens-threshold-frac: None
 dynamo:
-  hash: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
   install: true
+  wheel: "1.5.0.dev20260910"
 backend:
   type: sglang
   prefill_environment:
diff --git a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p6d-dep8-tp4-c45-mtp.yaml b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p6d-dep8-tp4-c45-mtp.yaml
index 7e305253ad..0597ee683f 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p6d-dep8-tp4-c45-mtp.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-1p6d-dep8-tp4-c45-mtp.yaml
@@ -8,10 +8,10 @@ identity:
     repo: nvidia/GLM-5.2-NVFP4
     revision: aec724e8c7b8ee9db3b48c01c320f63f9cdaf8aa
   container:
-    image: lmsysorg/sglang:nightly-dev-cu13-20260805-211ee642
+    image: lmsysorg/sglang:v0.5.19-cu130@sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9
   frameworks:
-    dynamo: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
-    sglang: nightly-dev-cu13-20260805-211ee642
+    dynamo: "1.5.0.dev20260910"
+    sglang: "0.5.19"
 resources:
   gpu_type: gb200
   gpus_per_node: 4
@@ -35,8 +35,8 @@ frontend:
     active-prefill-tokens-threshold: None
     active-prefill-tokens-threshold-frac: None
 dynamo:
-  hash: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
   install: true
+  wheel: "1.5.0.dev20260910"
 backend:
   type: sglang
   prefill_environment:
diff --git a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-2p1d-dep8-dep16-c128-mtp.yaml b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-2p1d-dep8-dep16-c128-mtp.yaml
index 76dd1d5794..ee55b1bd97 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-2p1d-dep8-dep16-c128-mtp.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5.2/gb200-fp4/agentic/disagg-gb200-2p1d-dep8-dep16-c128-mtp.yaml
@@ -8,10 +8,10 @@ identity:
     repo: nvidia/GLM-5.2-NVFP4
     revision: aec724e8c7b8ee9db3b48c01c320f63f9cdaf8aa
   container:
-    image: lmsysorg/sglang:nightly-dev-cu13-20260805-211ee642
+    image: lmsysorg/sglang:v0.5.19-cu130@sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9
   frameworks:
-    dynamo: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
-    sglang: nightly-dev-cu13-20260805-211ee642
+    dynamo: "1.5.0.dev20260910"
+    sglang: "0.5.19"
 resources:
   gpu_type: gb200
   gpus_per_node: 4
@@ -35,8 +35,8 @@ frontend:
     active-prefill-tokens-threshold: None
     active-prefill-tokens-threshold-frac: None
 dynamo:
-  hash: 71eb001e17fa73c742f0afe1a6ed96836cb135fd
   install: true
+  wheel: "1.5.0.dev20260910"
 backend:
   type: sglang
   prefill_environment:
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index 9eeeffd505..87e792e3ac 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -9193,7 +9193,7 @@ glm5.2-fp4-gb200-dynamo-sglang-agentic-disagg:
 # GLM-5.2 NVFP4 GB200 AgentX disaggregated variants use the committed
 # thinking-on golden acceptance length for two speculative tokens.
 glm5.2-fp4-gb200-dynamo-sglang-agentic-mtp:
-  image: lmsysorg/sglang:nightly-dev-cu13-20260805-211ee642
+  image: lmsysorg/sglang:v0.5.19-cu130@sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9
   model: nvidia/GLM-5.2-NVFP4
   model-prefix: glm5.2
   runner: cluster:gb200-nv
@@ -9201,7 +9201,7 @@ glm5.2-fp4-gb200-dynamo-sglang-agentic-mtp:
   framework: dynamo-sglang
   router:
     name: dynamo-router
-    version: 1.2.1
+    version: 1.5.0.dev20260910
   kv-p2p-transfer: nixl
   multinode: true
   disagg: true
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 8efc4d5c03..3bff152ca4 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -7179,3 +7179,11 @@
   description:
     - "Use the pinned SGLang nightly-dev-cu13-20260901-07c8f729 image with FlashInfer 0.6.18, which includes the BF16 TRTLLM MoE allocation fix for small-batch Blackwell execution. Model, TP8, HiCache, MTP settings and concurrency grid are unchanged."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2994
+
+- config-keys:
+    - glm5.2-fp4-gb200-dynamo-sglang-agentic-mtp
+  description:
+    - "Update the SGLang image from lmsysorg/sglang:nightly-dev-cu13-20260805-211ee642 (2026-08-05 cu13 dev nightly, build commit sgl-project/sglang@211ee642) to the digest-pinned v0.5.19 release lmsysorg/sglang:v0.5.19-cu130@sha256:d6e7288627be8b02be88e4bba38e73f6d50e2826869f753c13a4c4385ab3eda9 (build commit sgl-project/sglang@0bcd822377da7b5718e674eaf9c870d349424dd1, Docker Hub last pushed 2026-09-04T22:50:19Z)."
+    - "Move the three disaggregated recipes' Dynamo install from source commit 71eb001e17fa73c742f0afe1a6ed96836cb135fd to the ai-dynamo 1.5.0.dev20260910 wheel, which carries the ServerArgs.get_model_config()/use_mla_backend() compatibility shim for SGLang #36972 (ai-dynamo/dynamo#14234) and pins sglang==0.5.19 (ai-dynamo/dynamo#14151); align recipe identity.frameworks and the master router version to 1.5.0.dev20260910 / 0.5.19."
+    - "Topology, MTP settings, HiCache DRAM offload, NIXL KV transfer, concurrency 45/48/128 and NVIDIA/srt-slurm v1.0.50 are unchanged."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2992