diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml
index 5943109612..8187c94cd7 100644
--- a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml
@@ -2,7 +2,7 @@ name: "minimax-m3-vllm-disagg-b300-2p1d-dep2-dep4-fp4-8k1k-eagle3"
 
 model:
   path: "nvidia/MiniMax-M3-NVFP4"
-  container: "vllm/vllm-openai:nightly-8e981630c9336233ca9de91452f68918bddbc4e2"
+  container: "vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7"
   precision: fp4
 
 resources:
@@ -52,6 +52,7 @@ backend:
       kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
       attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
       block-size: 128
+      kv-cache-dtype: fp8
       gpu-memory-utilization: 0.95
       max-model-len: 9472
       language-model-only: true
@@ -70,8 +71,9 @@ backend:
       trust-remote-code: true
       no-enable-prefix-caching: true
       kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
-      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8", "minimax_m3_msa_decode_backend": "cutlass"}'
       block-size: 128
+      kv-cache-dtype: fp8
       gpu-memory-utilization: 0.9
       max-model-len: 9472
       language-model-only: true
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index c0426ea91b..48947fd70f 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -7872,7 +7872,7 @@ minimaxm3-fp4-b300-dynamo-vllm-mtp:
           dp-attn: false
 
 minimaxm3-fp4-b300-dynamo-vllm-mtp-legacy-dep4:
-  image: vllm/vllm-openai:nightly-8e981630c9336233ca9de91452f68918bddbc4e2
+  image: vllm/vllm-openai:nightly-5e35a6f4f9bbc217c599692157ca985c894373f7
   model: nvidia/MiniMax-M3-NVFP4
   model-prefix: minimaxm3
   runner: b300
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 75c8ce68c1..ac0e5d8e27 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -5396,3 +5396,10 @@
     - "Bump vLLM nightly image to nightly-5e35a6f4f9bbc217c599692157ca985c894373f7 (fixed-len MiniMax-M3 fix); update B300 FP4 MTP recipe YAML container fields to match (excluding 2p1d-dep2-dep4 used by legacy-dep4)"
     - "Enable VLLM_MINIMAX_M3_MSA_DECODE_BACKEND=cutlass in prefill and decode environments"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2471
+
+- config-keys:
+    - minimaxm3-fp4-b300-dynamo-vllm-mtp-legacy-dep4
+  description:
+    - "Bump vLLM nightly image to nightly-5e35a6f4f9bbc217c599692157ca985c894373f7 (fixed-len MiniMax-M3 fix); update 2p1d-dep2-dep4 recipe YAML container to match"
+    - "Enable minimax_m3_msa_decode_backend=cutlass in decode attention-config"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2483