diff --git a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/vllm/mi355x-fp4-mtp/agentic.yaml b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/vllm/mi355x-fp4-mtp/agentic.yaml
index e7a976b4b9..539f913fd0 100644
--- a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/vllm/mi355x-fp4-mtp/agentic.yaml
+++ b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/dsv4/vllm/mi355x-fp4-mtp/agentic.yaml
@@ -39,6 +39,7 @@ base:
         data-parallel-size: 1
         gpu-memory-utilization: 0.86
         moe-backend: aiter
+        monokernel: true
         compilation-config: '{"mode":3,"cudagraph_mode":"FULL_AND_PIECEWISE"}'
         # Six-token DSpark with probabilistic drafting. Throughput runs replace
         # verification with the golden acceptance length.
@@ -167,6 +168,8 @@ override_dep8_c64:
       args:
         tensor-parallel-size: 1
         data-parallel-size: 8
+        # Pro's monokernel requires TP8 without data parallelism.
+        monokernel: false
         prefill-schedule-interval: 8
         long-prefill-token-threshold: 16384
         max-num-seqs: 64
@@ -189,6 +192,7 @@ override_dep8_c128:
       args:
         tensor-parallel-size: 1
         data-parallel-size: 8
+        monokernel: false
         prefill-schedule-interval: 8
         long-prefill-token-threshold: 16384
         max-num-seqs: 128
@@ -211,6 +215,7 @@ override_dep8_c192:
       args:
         tensor-parallel-size: 1
         data-parallel-size: 8
+        monokernel: false
         prefill-schedule-interval: 8
         long-prefill-token-threshold: 16384
         max-num-seqs: 192
diff --git a/inferencex-e2e/configs/amd-master.yaml b/inferencex-e2e/configs/amd-master.yaml
index ed18bec2d6..5202840eb9 100644
--- a/inferencex-e2e/configs/amd-master.yaml
+++ b/inferencex-e2e/configs/amd-master.yaml
@@ -988,7 +988,7 @@ qwen3.5-fp8-mi325x-sglang-mtp:
 # DeepSeek-V4-Pro-0813 FP4 AgentX on one MI355X node using vLLM DSpark K6.
 # Keep the historical config key for changelog compatibility.
 dsv4-fp4-mi355x-vllm-agentic-mtp:
-  image: vllm/vllm-openai-rocm:nightly-e9757321527ca1ecd514c07c1418dd2c53da3d19
+  image: vllm/vllm-openai-rocm:nightly-e4a5b5537344b103ed10de7764378ef09604ef3d
   model: deepseek-ai/DeepSeek-V4-Pro-0813
   model-prefix: dsv4
   runner: cluster:mi355x-amds
diff --git a/inferencex-e2e/perf-changelog.yaml b/inferencex-e2e/perf-changelog.yaml
index 920eaaaf82..01ea97051f 100644
--- a/inferencex-e2e/perf-changelog.yaml
+++ b/inferencex-e2e/perf-changelog.yaml
@@ -11499,3 +11499,10 @@
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4509
   scenario-type: [agentic-coding]
   append-only: true
+
+- config-keys:
+    - dsv4-fp4-mi355x-vllm-agentic-mtp
+  description:
+    - "Update DeepSeek-V4-Pro-0813 on MI355X to vllm/vllm-openai-rocm:nightly-e4a5b5537344b103ed10de7764378ef09604ef3d and enable --monokernel for TP8 points; keep it disabled for data-parallel points."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4526
+  scenario-type: [agentic-coding]