diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml
index c1412b8f4..9677054b2 100644
--- a/.github/configs/nvidia-master.yaml
+++ b/.github/configs/nvidia-master.yaml
@@ -10293,3 +10293,216 @@ minimaxm2.5-fp4-gb200-dynamo-vllm:
           tp: 4
           ep: 4
           dp-attn: true
+
+minimaxm2.5-fp8-gb300-dynamo-vllm:
+  image: vllm/vllm-openai:v0.20.1
+  model: MiniMaxAI/MiniMax-M2.5
+  model-prefix: minimaxm2.5
+  runner: gb300-nv
+  precision: fp8
+  framework: dynamo-vllm
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - conc-list: [8, 16, 32, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [32, 64, 128, 256, 512]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [256, 512, 1024]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4ep.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [256, 512, 1024]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p1d-dep8.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+      - conc-list: [512, 1024, 2048]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - conc-list: [4096, 8192]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4-hi-conc.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - conc-list: [1024]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p3d-dep2.yaml"
+        decode:
+          num-worker: 3
+          tp: 2
+          ep: 2
+          dp-attn: true
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [16, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [256, 512]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep-hi-conc.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [32]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp2.yaml"
+        decode:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: false
+      - conc-list: [64, 128, 256, 512]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp4ep.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: false
+      - conc-list: [64]
+        prefill:
+          num-worker: 3
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-tp4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [256, 512]
+        prefill:
+          num-worker: 3
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - conc-list: [1024, 2048]
+        prefill:
+          num-worker: 3
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4-hi-conc.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - conc-list: [512, 1024, 2048]
+        prefill:
+          num-worker: 5
+          tp: 1
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-5p2d-dep4.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 4
+          dp-attn: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p1d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p1d-tp4.yaml
new file mode 100644
index 000000000..4475c4548
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p1d-tp4.yaml
@@ -0,0 +1,64 @@
+name: "minimax-m2.5-vllm-disagg-gb300-1p1d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "8x16x32x64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4.yaml
new file mode 100644
index 000000000..005d3ab45
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4.yaml
@@ -0,0 +1,69 @@
+name: "minimax-m2.5-vllm-disagg-gb300-1p2d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 1
+  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
+      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: "32x64x128x256x512"
+  # 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-fp8/1k1k/disagg-gb300-1p2d-tp4ep.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4ep.yaml
new file mode 100644
index 000000000..42e2bbff7
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-1p2d-tp4ep.yaml
@@ -0,0 +1,66 @@
+name: "minimax-m2.5-vllm-disagg-gb300-1p2d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 1
+  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
+      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: "256x512x1024"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p1d-dep8.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p1d-dep8.yaml
new file mode 100644
index 000000000..dadaea41c
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p1d-dep8.yaml
@@ -0,0 +1,83 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p1d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "256x512x1024"
+  # 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-fp8/1k1k/disagg-gb300-2p2d-dep4-hi-conc.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4-hi-conc.yaml
new file mode 100644
index 000000000..95a6f4032
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4-hi-conc.yaml
@@ -0,0 +1,82 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p2d-dep4-hi-conc"
+
+# 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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 2
+  gpus_per_prefill: 1
+  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
+      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"
+  # 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-fp8/1k1k/disagg-gb300-2p2d-dep4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4.yaml
new file mode 100644
index 000000000..90d14b5b0
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p2d-dep4.yaml
@@ -0,0 +1,82 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p2d-dep4"
+
+# 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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 2
+  gpus_per_prefill: 1
+  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
+      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: 1024
+  osl: 1024
+  concurrencies: "512x1024x2048"
+  # 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-fp8/1k1k/disagg-gb300-2p3d-dep2.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p3d-dep2.yaml
new file mode 100644
index 000000000..ef4bfc846
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/1k1k/disagg-gb300-2p3d-dep2.yaml
@@ -0,0 +1,68 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p3d-dep2"
+
+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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 3
+  gpus_per_prefill: 1
+  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
+      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: 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: 32
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "1024"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep-hi-conc.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep-hi-conc.yaml
new file mode 100644
index 000000000..f9e9ccf79
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep-hi-conc.yaml
@@ -0,0 +1,66 @@
+name: "minimax-m2.5-vllm-disagg-gb300-1p1d-tp4ep-hi-conc"
+
+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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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-fp8/8k1k/disagg-gb300-1p1d-tp4ep.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep.yaml
new file mode 100644
index 000000000..76e72c229
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-1p1d-tp4ep.yaml
@@ -0,0 +1,66 @@
+name: "minimax-m2.5-vllm-disagg-gb300-1p1d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "16x64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp2.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp2.yaml
new file mode 100644
index 000000000..f71458a70
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp2.yaml
@@ -0,0 +1,68 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p1d-tp2"
+
+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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: 2
+      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: "32"
+  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-fp8/8k1k/disagg-gb300-2p1d-tp4ep.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp4ep.yaml
new file mode 100644
index 000000000..668cf185b
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-2p1d-tp4ep.yaml
@@ -0,0 +1,66 @@
+name: "minimax-m2.5-vllm-disagg-gb300-2p1d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "64x128x256x512"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4-hi-conc.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4-hi-conc.yaml
new file mode 100644
index 000000000..94b866d95
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4-hi-conc.yaml
@@ -0,0 +1,82 @@
+name: "minimax-m2.5-vllm-disagg-gb300-3p1d-dep4-hi-conc"
+
+# 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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 3
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "1024x2048"
+  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-fp8/8k1k/disagg-gb300-3p1d-dep4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4.yaml
new file mode 100644
index 000000000..9bb6081db
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-dep4.yaml
@@ -0,0 +1,82 @@
+name: "minimax-m2.5-vllm-disagg-gb300-3p1d-dep4"
+
+# 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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 3
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "256x512"
+  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-fp8/8k1k/disagg-gb300-3p1d-tp4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-tp4.yaml
new file mode 100644
index 000000000..b638c0351
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-3p1d-tp4.yaml
@@ -0,0 +1,68 @@
+name: "minimax-m2.5-vllm-disagg-gb300-3p1d-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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 3
+  decode_workers: 1
+  gpus_per_prefill: 1
+  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
+      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: "64"
+  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-fp8/8k1k/disagg-gb300-5p2d-dep4.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-5p2d-dep4.yaml
new file mode 100644
index 000000000..ed2a9cdd4
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/8k1k/disagg-gb300-5p2d-dep4.yaml
@@ -0,0 +1,68 @@
+name: "minimax-m2.5-vllm-disagg-gb300-5p2d-dep4"
+
+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: "gb300"
+  gpus_per_node: 4
+  prefill_nodes: 2
+  decode_nodes: 2
+  prefill_workers: 5
+  decode_workers: 2
+  gpus_per_prefill: 1
+  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
+      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: "512x1024x2048"
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 1ce1698b8..1374a6492 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -3416,3 +3416,10 @@
     - "Add MiniMax-M2.5 FP8 B200 disaggregated multinode vLLM benchmarks via Dynamo"
     - "Add 1k1k/8k1k FP8 recipe set under benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-b200-fp8/"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1649
+
+- config-keys:
+    - minimaxm2.5-fp8-gb300-dynamo-vllm
+  description:
+    - "Add MiniMax-M2.5 FP8 GB300 disaggregated multinode vLLM benchmarks via Dynamo"
+    - "Add 1k1k/8k1k FP8 recipe set under benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8/"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1647
diff --git a/runners/launch_gb300-nv.sh b/runners/launch_gb300-nv.sh
index b47e103fd..4bffdb0b5 100644
--- a/runners/launch_gb300-nv.sh
+++ b/runners/launch_gb300-nv.sh
@@ -45,8 +45,11 @@ elif [[ $MODEL_PREFIX == "glm5" && $PRECISION == "fp4" ]]; then
 elif [[ $MODEL_PREFIX == "glm5" && $PRECISION == "fp8" ]]; then
     export MODEL_PATH=/scratch/models/GLM-5-FP8
     export SRT_SLURM_MODEL_PREFIX="glm-5-fp8"
+elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp8" ]]; then
+    export MODEL_PATH=/data/models/MiniMax-M2.5
+    export SRT_SLURM_MODEL_PREFIX="minimax-m2.5-fp8"
 else
-    echo "Unsupported model: $MODEL_PREFIX-$PRECISION. Supported models are: dsr1-fp4, dsr1-fp8, dsv4-fp4, glm5-fp4, glm5-fp8"
+    echo "Unsupported model: $MODEL_PREFIX-$PRECISION. Supported models are: dsr1-fp4, dsr1-fp8, dsv4-fp4, glm5-fp4, glm5-fp8, minimaxm2.5-fp8"
     exit 1
 fi
 
@@ -137,6 +140,12 @@ elif [[ $FRAMEWORK == "dynamo-sglang" && $MODEL_PREFIX == "glm5" ]]; then
     git checkout sa-submission-q2-2026
     mkdir -p recipes/sglang/glm5
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/sglang/glm5" recipes/sglang/glm5
+elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp8" ]]; then
+    git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
+    cd "$SRT_REPO_DIR"
+    git checkout main
+    mkdir -p recipes/vllm/minimax-m2.5-fp8
+    cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-fp8" recipes/vllm/minimax-m2.5-fp8
 else
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR"
@@ -370,8 +379,12 @@ if [[ "${RUN_EVAL:-false}" == "true" || "${EVAL_ONLY:-false}" == "true" ]]; then
         shopt -s nullglob
         for eval_file in "$EVAL_DIR"/*; do
             [ -f "$eval_file" ] || continue
-            cp "$eval_file" "$GITHUB_WORKSPACE/"
-            echo "Copied eval artifact: $(basename "$eval_file")"
+            eval_dest="$GITHUB_WORKSPACE/$(basename "$eval_file")"
+            if cp "$eval_file" "$eval_dest"; then
+                echo "Copied eval artifact: $(basename "$eval_file")"
+            else
+                echo "WARNING: Failed to copy eval artifact, continuing: $(basename "$eval_file")"
+            fi
         done
         shopt -u nullglob
     else