diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml
index b1ab08968..855ba0555 100644
--- a/.github/configs/nvidia-master.yaml
+++ b/.github/configs/nvidia-master.yaml
@@ -10080,6 +10080,141 @@ qwen3.5-fp8-h100-sglang-agentic:
       - { tp: 8, ep: 8, offloading: none,    conc-list: [1, 2, 4, 8, 12, 14, 16] }
       - { tp: 8, ep: 8, offloading: hicache, conc-list: [12, 14, 16, 20, 24, 28, 32, 42] }
 
+minimaxm2.5-fp8-gb200-dynamo-vllm:
+  image: vllm/vllm-openai:v0.20.1
+  model: MiniMaxAI/MiniMax-M2.5
+  model-prefix: minimaxm2.5
+  runner: gb200
+  precision: fp8
+  framework: dynamo-vllm
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - conc-list: [1, 4, 8, 16, 32, 64]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [2, 32, 64, 128, 256, 512]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [1024]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [512, 1024]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+      - conc-list: [4096]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml"
+        decode:
+          num-worker: 4
+          tp: 2
+          ep: 2
+          dp-attn: true
+      - conc-list: [4096, 8192]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 4
+          dp-attn: true
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [1, 4, 8, 16, 32, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [256, 512]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [1024, 2048, 4096]
+        prefill:
+          num-worker: 3
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: true
+
 
 minimaxm2.5-fp8-b200-dynamo-vllm:
   image: vllm/vllm-openai:v0.20.1
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml
new file mode 100644
index 000000000..120a35e45
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p1d-tp4.yaml
@@ -0,0 +1,67 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p1d-tp4"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "1x4x8x16x32x64"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml
new file mode 100644
index 000000000..6b5e76e42
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p2d-tp4.yaml
@@ -0,0 +1,67 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p2d-tp4"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "2x32x64x128x256x512"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml
new file mode 100644
index 000000000..765562d0c
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p3d-tp4ep.yaml
@@ -0,0 +1,72 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p3d-tp4ep"
+
+# Rate-matched tp4ep for FP8 GB200 1k/1k.
+# X_tp4ep_fp8_gb200 = 17.9k tok/s; P_per_worker = 48k; ideal X/P = 0.37; 1P:3D = 0.33 ✓
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 3
+  prefill_workers: 1
+  decode_workers: 3
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "1024"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml
new file mode 100644
index 000000000..aeeb8a012
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-1p4d-dep2.yaml
@@ -0,0 +1,74 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p4d-dep2"
+
+# Rate-matched dep2 for FP8 GB200 1k/1k.
+# X_dep2_fp8_gb200 = 12.7k tok/s; P_per_worker = 48k; ideal X/P = 0.27; 1P:4D = 0.25 ✓
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 4
+  gpus_per_prefill: 2
+  gpus_per_decode: 2
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 128
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 128
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "4096"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml
new file mode 100644
index 000000000..83bc7aeb2
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p1d-dep8.yaml
@@ -0,0 +1,86 @@
+name: "minimax-m2.5-vllm-disagg-gb200-2p1d-dep8"
+
+# model:
+#   path: "minimax-m2.5-fp8"
+#   container: "v0.18.1"
+#   precision: "fp8"
+
+# dynamo:
+#   version: 1.0.1
+#   install: true
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "512x1024"
+  # warmup_prompts: 1
+  # use_chat_template: false
+  # req_rate: "inf"
+  # random_range_ratio: 1.0
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml
new file mode 100644
index 000000000..534019222
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/1k1k/disagg-gb200-2p3d-dep4.yaml
@@ -0,0 +1,74 @@
+name: "minimax-m2.5-vllm-disagg-gb200-2p3d-dep4"
+
+# Rate-matched dep4 for FP8 GB200 1k/1k.
+# X_dep4_fp8_gb200 = 30.9k tok/s; P_per_worker = 48k; ideal X/P = 0.64; 2P:3D = 0.67 ✓
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 3
+  prefill_workers: 2
+  decode_workers: 3
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 128
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 128
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "4096x8192"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml
new file mode 100644
index 000000000..847c4b138
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4.yaml
@@ -0,0 +1,68 @@
+name: "minimax-m2.5-vllm-disagg-gb200-8k1k-1p1d-tp4"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      max-num-batched-tokens: 16384
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "1x4x8x16x32x64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml
new file mode 100644
index 000000000..61497185a
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-1p1d-tp4ep.yaml
@@ -0,0 +1,70 @@
+name: "minimax-m2.5-vllm-disagg-gb200-1p1d-tp4ep"
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      max-num-batched-tokens: 16384
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "256x512"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml
new file mode 100644
index 000000000..c92757146
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/8k1k/disagg-gb200-3p2d-dep4.yaml
@@ -0,0 +1,76 @@
+name: "minimax-m2.5-vllm-disagg-gb200-8k1k-3p2d-dep4"
+
+# Rate-matched dep4 for FP8 GB200 8k/1k.
+# X_dep4_fp8_gb200_8k ≈ 9.8k tok/s (from 5p2d-dep4 saturation);
+# P_per_worker_8k = 57k; ratio = X*8/P = 78.4/57 = 1.38; 3P:2D = 1.5 ✓ (closest int fit)
+
+model:
+  path: "minimax-m2.5-fp8"
+  container: "vllm/vllm-openai:v0.20.1"
+  precision: "fp8"
+
+dynamo:
+  install: true
+  wheel: "1.2.0.dev20260526"
+
+setup_script: install-deps.sh
+
+resources:
+  gpu_type: "gb200"
+  gpus_per_node: 4
+  prefill_nodes: 2
+  decode_nodes: 2
+  prefill_workers: 3
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  decode_environment:
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "mnnvl"
+
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-num-batched-tokens: 16384
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      safetensors-load-strategy: "prefetch"
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      stream-interval: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "1024x2048x4096"
diff --git a/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi355x_atom_mtp.sh b/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi355x_atom_mtp.sh
index 159598a07..2af10d749 100644
--- a/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi355x_atom_mtp.sh
+++ b/benchmarks/single_node/fixed_seq_len/dsr1_fp8_mi355x_atom_mtp.sh
@@ -36,7 +36,7 @@ if [ "$DP_ATTENTION" = "true" ]; then
     else #DP+TP
         PARALLEL_ARGS=(-tp "$TP" --enable-dp-attention )
     fi
-fi 
+fi
 
 SPEC_ARGS=(--method mtp --num-speculative-tokens 3 )
 
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 5bee49782..e28374a87 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -3444,3 +3444,10 @@
     - "Add MiniMax-M2.5 NVFP4 GB300 disaggregated multinode vLLM benchmarks via Dynamo"
     - "Add 1k1k/8k1k minimax recipe set under benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5/"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1641
+
+- config-keys:
+    - minimaxm2.5-fp8-gb200-dynamo-vllm
+  description:
+    - "Add MiniMax-M2.5 FP8 GB200 disaggregated multinode vLLM benchmarks via Dynamo"
+    - "Add 1k1k/8k1k FP8 recipe set under benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8/"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1648
diff --git a/runners/launch_gb200-nv.sh b/runners/launch_gb200-nv.sh
index dc7f83460..45ef3a952 100755
--- a/runners/launch_gb200-nv.sh
+++ b/runners/launch_gb200-nv.sh
@@ -57,8 +57,11 @@ elif [[ $FRAMEWORK == "dynamo-vllm" ]]; then
     elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp4" ]]; then
         export MODEL_PATH="/mnt/lustre01/models/MiniMax-M2.5-NVFP4"
         export SRT_SLURM_MODEL_PREFIX="minimax-m2.5-nvfp4"
+    elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp8" ]]; then
+        export MODEL_PATH="/mnt/lustre01/models/MiniMax-M2.5"
+        export SRT_SLURM_MODEL_PREFIX="minimax-m2.5-fp8"
     else
-        echo "Unsupported model prefix/precision combination: $MODEL_PREFIX/$PRECISION. Supported combinations for dynamo-vllm: kimik2.5/fp4, dsv4/fp4, minimaxm2.5/fp4"
+        echo "Unsupported model prefix/precision combination: $MODEL_PREFIX/$PRECISION. Supported combinations for dynamo-vllm: kimik2.5/fp4, dsv4/fp4, minimaxm2.5/fp4, minimaxm2.5/fp8"
         exit 1
     fi
 else
@@ -259,8 +262,16 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm2.5" ]]; then
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR" || exit 1
     cd "$SRT_REPO_DIR" || exit 1
     git checkout main || exit 1
-    mkdir -p recipes/vllm/minimax-m2.5-gb200 || exit 1
-    cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200" recipes/vllm/minimax-m2.5-gb200 || exit 1
+    if [[ $PRECISION == "fp8" ]]; then
+        mkdir -p recipes/vllm/minimax-m2.5-gb200-fp8 || exit 1
+        cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200-fp8" recipes/vllm/minimax-m2.5-gb200-fp8 || exit 1
+    elif [[ $PRECISION == "fp4" ]]; then
+        mkdir -p recipes/vllm/minimax-m2.5-gb200 || exit 1
+        cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-gb200" recipes/vllm/minimax-m2.5-gb200 || exit 1
+    else
+        echo "Unsupported minimaxm2.5 precision for GB200 dynamo-vllm: $PRECISION" >&2
+        exit 1
+    fi
 elif [[ $FRAMEWORK == "dynamo-vllm" ]]; then
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR"