diff --git a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/atom/mi355x-fp8-mtp/8k1k.yaml b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/atom/mi355x-fp8-mtp/8k1k.yaml
index 3c4ee34d62..5748844c78 100644
--- a/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/atom/mi355x-fp8-mtp/8k1k.yaml
+++ b/inferencex-e2e/benchmarks/single_node/srt-slurm-recipes/qwen3.5/atom/mi355x-fp8-mtp/8k1k.yaml
@@ -25,6 +25,7 @@ base:
         method: mtp
         num-speculative-tokens: 3
         trust-remote-code: true
+        no-enable_prefix_caching: true
       env:
         OMP_NUM_THREADS: '1'
   benchmark:
diff --git a/inferencex-e2e/configs/amd-master.yaml b/inferencex-e2e/configs/amd-master.yaml
index 2416001d97..715c297ab6 100644
--- a/inferencex-e2e/configs/amd-master.yaml
+++ b/inferencex-e2e/configs/amd-master.yaml
@@ -307,7 +307,7 @@ qwen3.5-fp8-mi355x-atom:
         srt-recipe: 8k1k.yaml
 
 qwen3.5-fp8-mi355x-atom-mtp:
-  image: rocm/atom:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom0.1.2.post
+  image: rocm/atom-dev:nightly_202610080328@sha256:9c3070bb7d2600d2e1bcfd03cf5f71e4dc0941f3e1d20d3a5845056aae2334eb
   model: Qwen/Qwen3.5-397B-A17B-FP8
   model-prefix: qwen3.5
   runner: mi355x
diff --git a/inferencex-e2e/perf-changelog.yaml b/inferencex-e2e/perf-changelog.yaml
index b4779bb289..b33c31f2e0 100644
--- a/inferencex-e2e/perf-changelog.yaml
+++ b/inferencex-e2e/perf-changelog.yaml
@@ -10351,3 +10351,9 @@
   description:
     - "Update SGLang image from v0.5.12-cu130 to v0.5.21-cu130 and rename cuda-graph-max-bs to cuda-graph-max-bs-decode."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3868
+
+- config-keys:
+    - qwen3.5-fp8-mi355x-atom-mtp
+  description:
+    - "Update ATOM to v0.1.7 (rocm/atom-dev:nightly_202610080328) and keep prefix caching disabled as in the old image."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3918