diff --git a/benchmarks/single_node/agentic/dsv4_fp4_mi355x_sglang_mtp.sh b/benchmarks/single_node/agentic/dsv4_fp4_mi355x_sglang_mtp.sh
index 61add3eaf..2a32f7c9b 100644
--- a/benchmarks/single_node/agentic/dsv4_fp4_mi355x_sglang_mtp.sh
+++ b/benchmarks/single_node/agentic/dsv4_fp4_mi355x_sglang_mtp.sh
@@ -73,6 +73,9 @@ export SGLANG_DSV4_REASONING_EFFORT=high
 export SGLANG_USE_ROCM700A=0
 export SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton
 export AITER_BF16_FP8_MOE_BOUND=0
+# aiter batched GEMM for the absorbed MLA projections, carried by the v0.5.18
+# image and off by default in environ.py.
+export SGLANG_OPT_USE_AITER_BATCHED_GEMM=1
 
 # Unified radix tree: per-component (full-attn / SWA) cache management for
 # hybrid-attention models, plus proactive release of out-of-window SWA KV
@@ -130,8 +133,10 @@ SGLANG_BACKEND_PORT="$PORT"
 # cap; same value the multi-node DeepSeek-V4-Pro-AgentX no_dp profile uses.
 CHUNKED_PREFILL_SIZE=8192
 # MTP adds a draft KV pool and extra graph captures on top of the spec-none
-# footprint, which ran at 0.90.
-MEM_FRACTION_STATIC=0.85
+# footprint, which ran at 0.90. 0.89 recovers most of that: the DSv4 compressor
+# state pools are sized from the full-attention pool and allocated after it,
+# outside this budget, so the remainder has to stay large enough to cover them.
+MEM_FRACTION_STATIC=0.89
 PARALLEL_ARGS=(--tensor-parallel-size "$TP")
 if [ "$DP_ATTENTION" = "true" ]; then
     USE_SGLANG_ROUTER=true
@@ -213,7 +218,7 @@ SGLANG_CMD=(
     --page-size 256
     --swa-full-tokens-ratio 0.10
     --kv-cache-dtype fp8_e4m3
-    --disable-shared-experts-fusion
+    --enforce-shared-experts-fusion
     --tool-call-parser deepseekv4
     --reasoning-parser deepseek-v4
     --chunked-prefill-size "$CHUNKED_PREFILL_SIZE"
diff --git a/configs/amd-master.yaml b/configs/amd-master.yaml
index 67a060539..2195bad7e 100644
--- a/configs/amd-master.yaml
+++ b/configs/amd-master.yaml
@@ -1716,18 +1716,9 @@ glm5.2-fp4-mi355x-atom-agentic-mtp:
        - { tp: 8, kv-offloading: none, conc-list: [1, 2, 4], spec-decoding: mtp }
 
 
-# DeepSeek-V4-Pro FP4 agentic-coding benchmark on MI355X via SGLang with the
-# in-checkpoint MTP head (EAGLE path, depth 3 -> golden AL 2.49 for
-# thinking_on, golden_al_distribution/dsv4_mtp.yaml). Restores the single-node
-# SGLang AgentX coverage removed in de493d859 (PR #2531, which deleted the
-# spec-none dsv4-fp4-mi355x-sglang-agentic-hicache key and its script) so the
-# aggregated SGLang path is comparable with dsv4-fp4-mi355x-vllm-agentic-mtp
-# and dsv4-fp4-mi355x-sglang-disagg-agentic-hicache-mtp. The image matches the
-# already-green disagg AgentX entry. Pure TP8 only: DSA + dp-attention hangs a
-# collective under long-context prefill, so no DEP arm ships until that path is
-# validated. conc 16 appears on both arms to isolate the host KV tier's gain.
+
 dsv4-fp4-mi355x-sglang-agentic-mtp:
-  image: lmsysorg/sglang-rocm:v0.5.17-rocm720-mi35x-20260813
+  image: lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260822
   model: deepseek-ai/DeepSeek-V4-Pro
   model-prefix: dsv4
   runner: cluster:mi355x-amds
@@ -1738,5 +1729,5 @@ dsv4-fp4-mi355x-sglang-agentic-mtp:
     agentic-coding:
     - dram-utilization: 0.80
       search-space:
-      - { tp: 8, ep: 1, dp-attn: false, kv-offloading: none, conc-list: [1, 2, 4, 8, 16], spec-decoding: mtp }
+      - { tp: 4, ep: 1, dp-attn: false, kv-offloading: none, conc-list: [1, 2, 4, 8, 10], spec-decoding: mtp }
       - { tp: 8, ep: 1, dp-attn: false, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [16, 32, 48], spec-decoding: mtp }
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index a9d2492a0..877f87b74 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -6387,3 +6387,15 @@
     - "Replace the internal TileRT 0.1.5.post2+inferencex.1 queueing backport with the official 0.1.5.post3 PyPI wheel now that upstream has published it; remove the internal wheel builder and router patch"
     - "Bump the shared TileRT dependency and router metadata to 0.1.5.post3 for both fixed-seq-length and AgentX configurations while retaining the AgentX 1800-second queue timeout"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2650
+
+- config-keys:
+    - dsv4-fp4-mi355x-sglang-agentic-mtp
+  scenario-type:
+    - agentic-coding
+  description:
+    - "Update image from lmsysorg/sglang-rocm:v0.5.17-rocm720-mi35x-20260813 to lmsysorg/sglang-rocm:v0.5.18-rocm720-mi35x-20260822."
+    - "Enable SGLANG_OPT_USE_AITER_BATCHED_GEMM=1, the aiter batched GEMM for the absorbed MLA projections, which defaults off."
+    - "Raise mem-fraction-static from 0.85 to 0.89, enlarging the full-attention KV pool. The DSv4 compressor state pools are sized from that pool and allocated after it, outside this budget, so the remainder still has to cover them."
+    - "Switch --disable-shared-experts-fusion to --enforce-shared-experts-fusion, which fuses the single shared expert into the routed list. The tuned MoE tables in this image cover the resulting (inter_dim 384, expert 385, topk 7) shape at the token sizes this workload hits."
+    - "Split the parallelism by concurrency: add a TP4 arm at [1, 2, 4, 8, 10] with no host KV tier, where halving the GPUs per replica raises throughput per GPU, and remove the TP8 no-offload arm [1, 2, 4, 8, 16]. This drops the published TP8 concurrency 1, 2, 4, 8 and 16 points and adds TP4 concurrency 1, 2, 4, 8 and 10."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2713