diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p1d-dep2-dep4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p1d-dep2-dep4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..c239a1b98c
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p1d-dep2-dep4-eagle3-8k1k.yaml
@@ -0,0 +1,93 @@
+name: "minimax-m3-vllm-disagg-b300-1p1d-dep2-dep4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: "fp4"
+
+resources:
+  gpu_type: "b300"
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 0
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 2
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: "1.3.0.dev20260710"
+
+frontend:
+  type: "dynamo"
+  enable_multiple_frontends: false
+
+backend:
+  type: "vllm"
+  connector:
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "trtllm"
+    UCX_TLS: "cuda_copy,cuda_ipc,rc"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: "high"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "trtllm"
+    UCX_TLS: "cuda_copy,cuda_ipc,rc"
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.9
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  req_rate: "inf"
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "64x128x256"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p2d-tp4-tp4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p2d-tp4-tp4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..a6dc5f9892
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p2d-tp4-tp4-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p2d-tp4-tp4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 2
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "4x64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp2-tp2-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp2-tp2-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..55f620f09d
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp2-tp2-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p3d-tp2-tp2-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 0
+  prefill_workers: 1
+  decode_workers: 3
+  gpus_per_prefill: 2
+  gpus_per_decode: 2
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp4-tp4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp4-tp4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..9d5425c8f3
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp4-tp4-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p3d-tp4-tp4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 3
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "4x48x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p4d-tp2-tp2-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p4d-tp2-tp2-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..16c2d18234
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p4d-tp2-tp2-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p4d-tp2-tp2-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 4
+  gpus_per_prefill: 2
+  gpus_per_decode: 2
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "64"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p5d-tp4-tp4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p5d-tp4-tp4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..db76ffabfa
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p5d-tp4-tp4-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p5d-tp4-tp4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 5
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "40x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp2-tp2-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp2-tp2-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..cc92ae5ca0
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp2-tp2-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p7d-tp2-tp2-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 7
+  gpus_per_prefill: 2
+  gpus_per_decode: 2
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 2
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "64x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp4-tp4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp4-tp4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..5453387a9f
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp4-tp4-eagle3-8k1k.yaml
@@ -0,0 +1,90 @@
+name: "minimax-m3-vllm-disagg-b300-1p7d-tp4-tp4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: fp4
+
+resources:
+  gpu_type: b300
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 3
+  prefill_workers: 1
+  decode_workers: 7
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: cuda_copy,cuda_ipc,rc
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      enable-expert-parallel: false
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  req_rate: inf
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
+  concurrencies: "16x28x128"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml
new file mode 100644
index 0000000000..5ac95488b9
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml
@@ -0,0 +1,95 @@
+name: "minimax-m3-vllm-disagg-b300-2p1d-dep2-dep4-fp4-8k1k-eagle3"
+model:
+  path: "nvidia/MiniMax-M3-NVFP4"
+  container: "vllm/vllm-openai:nightly"
+  precision: "fp4"
+
+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: 4
+
+dynamo:
+  install: true
+  version: 1.3.0.dev20260710
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+
+backend:
+  type: vllm
+  connector: null
+  allow_prefill_decode_colocation: true
+  allow_prefill_decode_colocation_across_nodes: true
+
+  prefill_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: "cuda_copy,cuda_ipc,rc"
+
+  decode_environment:
+    VLLM_FLOAT32_MATMUL_PRECISION: high
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: trtllm
+    UCX_TLS: "cuda_copy,cuda_ipc,rc"
+
+  vllm_config:
+    prefill:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 2
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.95
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":1,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-cudagraph-capture-size: 2048
+      max-num-batched-tokens: 16384
+
+    decode:
+      no-enable-flashinfer-autotune: true
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      attention-config: '{"backend": "FLASHINFER", "use_trtllm_attention": true, "indexer_kv_dtype": "fp8"}'
+      block-size: 128
+      gpu-memory-utilization: 0.9
+      max-model-len: 9472
+      language-model-only: true
+      speculative-config: '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}'
+      stream-interval: 32
+      max-num-seqs: 1024
+      max-num-batched-tokens: 16384
+      max-cudagraph-capture-size: 2048
+
+health_check:
+  max_attempts: 360
+  interval_seconds: 10
+
+benchmark:
+  type: "sa-bench"
+  isl: 8192
+  osl: 1024
+  concurrencies: "128x256"
+  req_rate: "inf"
+  num_warmup_mult: 0
+  random_range_ratio: 0.8
+  use_chat_template: true
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index f90c510186..b6d4fdc81e 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -13163,6 +13163,151 @@ minimaxm3-fp4-b300-dynamo-vllm:
           ep: 8
           dp-attn: false
 
+minimaxm3-fp4-b300-dynamo-vllm-mtp:
+  image: vllm/vllm-openai:nightly
+  model: nvidia/MiniMax-M3-NVFP4
+  model-prefix: minimaxm3
+  runner: b300
+  precision: fp4
+  framework: dynamo-vllm
+  router: { name: dynamo-router, version: "1.3.0.dev20260710" }
+  kv-p2p-transfer: nixl
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - spec-decoding: "mtp"
+        conc-list: [4, 48, 128]
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp4-tp4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 3
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - spec-decoding: "mtp"
+        conc-list: [4, 64, 128]
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p2d-tp4-tp4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 2
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - spec-decoding: "mtp"
+        conc-list: [16, 28, 128]
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp4-tp4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 7
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - spec-decoding: "mtp"
+        conc-list: [40, 128]
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p5d-tp4-tp4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 5
+          tp: 4
+          ep: 1
+          dp-attn: false
+
+      - spec-decoding: "mtp"
+        conc-list: [64, 128]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p7d-tp2-tp2-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 7
+          tp: 2
+          ep: 1
+          dp-attn: false
+      - spec-decoding: "mtp"
+        conc-list: [64]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p4d-tp2-tp2-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 4
+          tp: 2
+          ep: 1
+          dp-attn: false
+      - spec-decoding: "mtp"
+        conc-list: [64, 128]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p3d-tp2-tp2-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 3
+          tp: 2
+          ep: 1
+          dp-attn: false
+
+      - spec-decoding: "mtp"
+        conc-list: [64, 128, 256]
+        prefill:
+          num-worker: 1
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/1p1d-dep2-dep4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+      - spec-decoding: "mtp"
+        conc-list: [128, 256]
+        prefill:
+          num-worker: 2
+          tp: 2
+          ep: 2
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/minimax-m3/b300-fp4/8k1k/mtp/2p1d-dep2-dep4-eagle3-8k1k.yaml"
+        decode:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+
 # MiniMax-M3 GB300 disagg sweep — refreshed recipe set (no Marlin variants).
 # All prefill DEP2 (TP1 DP2 EP, 2 GPU/worker). Decode: DEP4, TEP8, DEP8, TEP4.
 # 4 GPU/node (GB300 NVL72). kv-cache-dtype=fp8. srun_options mem=0 required.
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 57d510dd15..f5807a318d 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -4750,3 +4750,9 @@
     - "Image: lmsysorg/sglang:nightly-dev-cu13-20260709-074bb928"
     - "6 topologies across 1k/1k and 8k/1k: 1P1D TP4 STP + wide-EP (DEP4 prefill / DEP16 decode) from 1P1D up to 8P1D, recipes under benchmarks/multi_node/srt-slurm-recipes/sglang/qwen3.5/gb300-fp8/"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2137
+
+- config-keys:
+    - minimaxm3-fp4-b300-dynamo-vllm-mtp
+  description:
+    - "Add MiniMax M3 NVFP4 B300 Dynamo vLLM disaggregated EAGLE3 Pareto configurations for the 8k/1k workload"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2182
diff --git a/runners/launch_b300-nv.sh b/runners/launch_b300-nv.sh
index 6810ee5d85..d5bc243e90 100644
--- a/runners/launch_b300-nv.sh
+++ b/runners/launch_b300-nv.sh
@@ -76,15 +76,18 @@ elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "dsv4" ]]; then
     git checkout aflowers/vllm-gb200-v0.20.0
     mkdir -p recipes/vllm/deepseek-v4
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4" recipes/vllm/deepseek-v4
-elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && ( $PRECISION == "fp4" || $PRECISION == "fp8" ) ]]; then
+elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && $PRECISION == "fp4" ]]; then
+    git clone --branch main --single-branch https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
+    cd "$SRT_REPO_DIR" || exit 1
+    mkdir -p recipes/vllm/minimax-m3
+    cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/minimax-m3" recipes/vllm/minimax-m3
+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
-    if [[ $PRECISION == "fp8" ]]; then
-        SRTCTL_SETUP_SCRIPT="minimax-m3-vllm-fixes.sh"
-    fi
+    SRTCTL_SETUP_SCRIPT="minimax-m3-vllm-fixes.sh"
     # NVIDIA/srt-slurm#38
     git show 22d46ba9971615016d2339c9ffbc7b4597accfad --format= -- src/srtctl/core/ip_utils/get_node_ip.sh | git apply - || exit 1
     if [[ -n "$SRTCTL_SETUP_SCRIPT" ]]; then
@@ -181,6 +184,9 @@ SRTCTL_APPLY_ARGS=(
     -f "$CONFIG_FILE"
     --tags "b300,${MODEL_PREFIX},${PRECISION},${ISL}x${OSL},infmax-$(date +%Y%m%d)"
 )
+if [[ "$MODEL_PREFIX" == "minimaxm3" && "$PRECISION" == "fp4" ]]; then
+    SRTCTL_APPLY_ARGS+=(--no-preflight)
+fi
 if [[ -n "$SRTCTL_SETUP_SCRIPT" ]]; then
     SRTCTL_APPLY_ARGS+=(--setup-script "$SRTCTL_SETUP_SCRIPT")
 fi