diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml
index c8fe1ea97..9bc22043d 100644
--- a/.github/configs/nvidia-master.yaml
+++ b/.github/configs/nvidia-master.yaml
@@ -12363,6 +12363,210 @@ minimaxm2.5-fp8-b300-dynamo-vllm:
           ep: 4
           dp-attn: true
 
+# MiniMax-M3 B300 disaggregated vLLM recipes sourced from
+# NVIDIA/srt-slurm#223 at 5caabe364e1ef531fab9926c75e32ae8927b1553.
+# The upstream recipes use DEP2 prefill workers and a mix of TEP8, DEP8,
+# and DEP4 decode workers across the submitted 1k1k/8k1k frontier.
+minimaxm3-fp8-b300-dynamo-vllm:
+  image: vllm/vllm-openai:minimax-m3
+  model: MiniMaxAI/MiniMax-M3-MXFP8
+  model-prefix: minimaxm3
+  runner: b300
+  precision: fp8
+  framework: dynamo-vllm
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 1024
+      osl: 1024
+      search-space:
+      - conc-list: [4, 16, 64, 128, 4096]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/1p1d-dep2-tep8-1k1k.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: false
+      - conc-list: [2048]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/1p2d-dep2-dep4-1k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: true
+      - conc-list: [512, 4096]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-dep8-1k1k.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: true
+      - conc-list: [32]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-tep8-1k1k.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: false
+      - conc-list: [16]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/2p2d-dep2-tep8-1k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: false
+      - conc-list: [4]
+        prefill:
+          num-worker: 3
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/1k1k/3p2d-dep2-tep8-1k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: false
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [128]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-dep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 1
+          dp-attn: true
+      - conc-list: [256, 512]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 1
+          dp-attn: true
+      - conc-list: [16]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: false
+      - conc-list: [512]
+        prefill:
+          num-worker: 3
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 1
+          dp-attn: true
+      - conc-list: [32]
+        prefill:
+          num-worker: 3
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-tep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: false
+      - conc-list: [4096]
+        prefill:
+          num-worker: 4
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 1
+          dp-attn: true
+      - conc-list: [4096]
+        prefill:
+          num-worker: 4
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/4p3d-dep2-dep4-8k1k.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 1
+          dp-attn: true
+      - conc-list: [4, 64]
+        prefill:
+          num-worker: 5
+          tp: 2
+          ep: 1
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp8/8k1k/5p2d-dep2-tep8-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 8
+          ep: 8
+          dp-attn: false
+
 minimaxm2.5-fp8-gb300-dynamo-vllm:
   image: vllm/vllm-openai:v0.20.1
   model: MiniMaxAI/MiniMax-M2.5
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p1d-dep2-tep8-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p1d-dep2-tep8-1k1k.yaml
new file mode 100644
index 000000000..7f93d1aa1
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p1d-dep2-tep8-1k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-1p1d-fp8-dep2-tep8-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 4096
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "4x16x64x128x4096"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p2d-dep2-dep4-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p2d-dep2-dep4-1k1k.yaml
new file mode 100644
index 000000000..8dc458eab
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/1p2d-dep2-dep4-1k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-1p2d-fp8-dep2-dep4-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "2048"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-dep8-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-dep8-1k1k.yaml
new file mode 100644
index 000000000..afe1c3e49
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-dep8-1k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-2p1d-fp8-dep2-dep8-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 4096
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "512x4096"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-tep8-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-tep8-1k1k.yaml
new file mode 100644
index 000000000..7a2d87740
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p1d-dep2-tep8-1k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-2p1d-fp8-dep2-tep8-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 4096
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "32"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p2d-dep2-tep8-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p2d-dep2-tep8-1k1k.yaml
new file mode 100644
index 000000000..33435b328
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/2p2d-dep2-tep8-1k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-2p2d-fp8-dep2-tep8-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 4096
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "16"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/3p2d-dep2-tep8-1k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/3p2d-dep2-tep8-1k1k.yaml
new file mode 100644
index 000000000..ac98b9cc8
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/1k1k/3p2d-dep2-tep8-1k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-3p2d-fp8-dep2-tep8-1k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 3
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  decode_environment:
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 2048
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 2304
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 4096
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 8196
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 1024
+  osl: 1024
+  concurrencies: "4"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-dep8-8k1k.yaml
new file mode 100644
index 000000000..cb9256e25
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/1p2d-dep2-dep8-8k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-1p2d-fp8-dep2-dep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "128"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml
new file mode 100644
index 000000000..7a805df7f
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-dep8-8k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-2p2d-fp8-dep2-dep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "256x512"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
new file mode 100644
index 000000000..8dc4c5c06
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/2p2d-dep2-tep8-8k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-2p2d-fp8-dep2-tep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 2
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "16"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml
new file mode 100644
index 000000000..8ff86e9d9
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-dep8-8k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-3p2d-fp8-dep2-dep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 3
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "512"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-tep8-8k1k.yaml
new file mode 100644
index 000000000..a90bfcf32
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/3p2d-dep2-tep8-8k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-3p2d-fp8-dep2-tep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 3
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "32"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml
new file mode 100644
index 000000000..8feb79b71
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p2d-dep2-dep8-8k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-4p2d-fp8-dep2-dep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 4
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "4096"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p3d-dep2-dep4-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p3d-dep2-dep4-8k1k.yaml
new file mode 100644
index 000000000..5639c2eda
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/4p3d-dep2-dep4-8k1k.yaml
@@ -0,0 +1,81 @@
+name: "minimax-m3-vllm-disagg-b300-4p3d-fp8-dep2-dep4-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 4
+  decode_workers: 3
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 512
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "4096"
+  req_rate: "inf"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/5p2d-dep2-tep8-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/5p2d-dep2-tep8-8k1k.yaml
new file mode 100644
index 000000000..1f517d815
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp8/8k1k/5p2d-dep2-tep8-8k1k.yaml
@@ -0,0 +1,80 @@
+name: "minimax-m3-vllm-disagg-b300-5p2d-fp8-dep2-tep8-8k1k"
+
+model:
+  path: "MiniMaxAI/MiniMax-M3-MXFP8"
+  container: "vllm/vllm-openai:minimax-m3"
+  precision: "fp8"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 2
+  decode_nodes: 2
+  prefill_workers: 5
+  decode_workers: 2
+  gpus_per_prefill: 2
+  gpus_per_decode: 8
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260614
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    UCX_NET_DEVICES: "all"
+    UCX_TLS: "rc,cuda_ipc,cuda_copy,sm,self,tcp"
+
+  vllm_config:
+    prefill:
+      tensor-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      tensor-parallel-size: 8
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      block-size: 128
+      gpu-memory-utilization: 0.90
+      max-model-len: 9472
+      language-model-only: true
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 4096
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "4x64"
+  req_rate: "inf"
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 67a91d5d2..842a98517 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -3887,3 +3887,12 @@
     - "Align MiniMax-M3 B300 vLLM fixed-sequence serving with MiniMax-M2.5 FP8 settings by setting VLLM_FLOAT32_MATMUL_PRECISION=high and restoring max cudagraph capture size 2048."
     - "Add TP4+EP4 coverage for MiniMax-M3 B300: DP-attention rows for 1k1k/8k1k and the missing non-DP-attention row for 8k1k."
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1781
+
+- config-keys:
+    - minimaxm3-fp8-b300-dynamo-vllm
+  description:
+    - "Add MiniMax-M3 MXFP8 B300 Dynamo vLLM disaggregated 1k1k and 8k1k benchmark coverage"
+    - "Import fourteen STP recipes from NVIDIA/srt-slurm PR #223 at 5caabe364e1ef531fab9926c75e32ae8927b1553"
+    - "Cover DEP2 prefill with TEP8, DEP8, and DEP4 decode topologies across submitted concurrencies 4 through 4096"
+    - "Image: vllm/vllm-openai:minimax-m3; model: MiniMaxAI/MiniMax-M3-MXFP8"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1788
diff --git a/runners/launch_b300-nv.sh b/runners/launch_b300-nv.sh
index a941860c0..f2a83e4b3 100644
--- a/runners/launch_b300-nv.sh
+++ b/runners/launch_b300-nv.sh
@@ -45,8 +45,11 @@ elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp4" && $FRAMEWORK == "
 elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp8" && $FRAMEWORK == "dynamo-vllm" ]]; then
     export MODEL_PATH="/data/models/MiniMax-M2.5"
     export SRT_SLURM_MODEL_PREFIX="minimax-m2.5-fp8"
+elif [[ $MODEL_PREFIX == "minimaxm3" && $PRECISION == "fp8" && $FRAMEWORK == "dynamo-vllm" ]]; then
+    export MODEL_PATH="/data/models/MiniMax-M3-MXFP8"
+    export SRT_SLURM_MODEL_PREFIX="MiniMaxAI/MiniMax-M3-MXFP8"
 else
-    echo "Unsupported model: $MODEL_PREFIX-$PRECISION. Supported models are: dsr1-fp4, dsr1-fp8, dsv4-fp4 with dynamo-vllm, minimaxm2.5-fp4 with dynamo-vllm, minimaxm2.5-fp8 with dynamo-vllm"
+    echo "Unsupported model: $MODEL_PREFIX-$PRECISION. Supported models are: dsr1-fp4, dsr1-fp8, dsv4-fp4 with dynamo-vllm, minimaxm2.5-fp4 with dynamo-vllm, minimaxm2.5-fp8 with dynamo-vllm, minimaxm3-fp8 with dynamo-vllm"
     exit 1
 fi
 
@@ -79,6 +82,12 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm2.5" && $PRECIS
     git checkout main
     mkdir -p recipes/vllm/minimax-m2.5-fp8
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m2.5-b300-fp8" recipes/vllm/minimax-m2.5-fp8
+elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && $PRECISION == "fp8" ]]; then
+    git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
+    cd "$SRT_REPO_DIR" || exit 1
+    git checkout sa-submission-q2-2026
+    mkdir -p recipes/vllm/minimax-m3
+    cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3" recipes/vllm/minimax-m3
 else
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR" || exit 1