diff --git a/benchmarks/benchmark_lib.sh b/benchmarks/benchmark_lib.sh
index bda03e8b6e..a41936ea18 100644
--- a/benchmarks/benchmark_lib.sh
+++ b/benchmarks/benchmark_lib.sh
@@ -1797,7 +1797,11 @@ build_replay_cmd() {
     # X-Correlation-ID is useful tracing metadata but does not establish that
     # binding by itself. AIPerf emits nvext.session_control bind/close actions
     # keyed by the stable conversation correlation ID when this flag is set.
-    if [[ "${FRAMEWORK:-}" == dynamo-* ]]; then
+    # Opt-out: recipes set AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING=0 to skip this.
+    # aiperf's conv-aware routing emits nvext.session_control, a removed POC field
+    # (dynamo #9920 / v1.3.0-dev) that current dynamo builds reject with a 400
+    # (they moved to router/routing_constraints/agent_context). Default stays on.
+    if [[ "${FRAMEWORK:-}" == dynamo-* && "${AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING:-1}" != "0" ]]; then
         REPLAY_CMD+=" --use-dynamo-conv-aware-routing"
         # The upstream 300s affinity TTL is shorter than an overloaded
         # high-concurrency agentic request. Keep bindings alive across long
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp4-mtp-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp4-mtp-agentic.yaml
new file mode 100644
index 0000000000..6d1681793b
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp4-mtp-agentic.yaml
@@ -0,0 +1,147 @@
+name: "svf-vllm-agg-gb300-tp4-mtp-agentic"
+
+# GB300 AgentX aggregate topology: one TP4 worker occupies one four-GPU node
+# and serves both prefill and decode at concurrency 4. Scheduler, CUDA-graph,
+# and memory settings match the B300 vLLM TP4 MTP agentic configuration.
+
+model:
+  path: "deepseek-v4-pro"
+  container: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro"
+  container:
+    image: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  frameworks:
+    dynamo: "1.2.1"
+
+dynamo:
+  wheel: "1.2.1"
+  install: true
+
+environment:
+  DYNAMO_WHEEL_DIRS: "/srtctl-wheels"
+  ETCD_LEASE_TTL: "7200"
+
+setup_script: vllm-container-deps.sh
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 2160
+  interval_seconds: 10
+
+resources:
+  gpu_type: "gb300"
+  gpus_per_node: 4
+  agg_nodes: 1
+  agg_workers: 1
+  gpus_per_agg: 4
+
+infra:
+  etcd_nats_dedicated_node: false
+  nats_max_payload_mb: 32
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+  args:
+    router-mode: "kv"
+    router-reset-states: true
+    router-temperature: 0.0
+    router-queue-threshold: 65536
+    active-decode-blocks-threshold: "None"
+    active-prefill-tokens-threshold: "None"
+    active-prefill-tokens-threshold-frac: "None"
+    tokenizer: "fastokens"
+
+backend:
+  type: vllm
+  connector: null
+  mooncake_kv_store:
+    store_config:
+      metadata_server: "P2PHANDSHAKE"
+      global_segment_size: "150GB"
+      local_buffer_size: "4GB"
+      protocol: "rdma"
+      device_name: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+      mode: "embedded"
+      enable_offload: false
+  aggregated_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "16"
+    TILELANG_CLEANUP_TEMP_FILES: "1"
+    VLLM_USE_NCCL_SYMM_MEM: "0"
+    TORCH_SYMMMEM: "NVSHMEM"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_USE_RUST_FRONTEND: "1"
+    VLLM_ALLREDUCE_USE_FLASHINFER: "1"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "auto"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_SPARSE_INDEXER_MAX_LOGITS_MB: "1024"
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: "1800"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    NCCL_P2P_LEVEL: "NVL"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-tp4-mtp-{job_id}"
+  vllm_config:
+    aggregated:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      disable-custom-all-reduce: true
+      enable-cumem-allocator: true
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      max-model-len: 1048576
+      max-num-seqs: 16
+      max-num-batched-tokens: 8192
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","cudagraph_capture_sizes":[4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64],"mode":0}'
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      gpu-memory-utilization: 0.93
+      stream-interval: 10
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+
+sbatch_directives:
+  cpus-per-task: "72"
+  mem: "0"
+
+srun_options:
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: "/infmax-workspace"
+    RESULT_DIR: "/logs/agentic"
+    PORT: "8000"
+    IS_MULTINODE: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
+    AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
+    HF_HUB_CACHE: "/hf_hub_cache"
+    WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp8-mtp-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp8-mtp-agentic.yaml
new file mode 100644
index 0000000000..06b916659f
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/agg-gb300-tp8-mtp-agentic.yaml
@@ -0,0 +1,150 @@
+name: "svf-vllm-agg-gb300-tp8-mtp-agentic"
+
+# Validated GB300 AgentX aggregate topology: one TP8 worker spans two
+# four-GPU nodes and serves both prefill and decode at concurrency 1.
+
+model:
+  path: "deepseek-v4-pro"
+  container: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro"
+  container:
+    image: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  frameworks:
+    dynamo: "1.2.1"
+
+dynamo:
+  wheel: "1.2.1"
+  install: true
+
+environment:
+  DYNAMO_WHEEL_DIRS: "/srtctl-wheels"
+  # The frontend shares Grace CPU capacity with the long TP8 cold start.
+  ETCD_LEASE_TTL: "7200"
+
+setup_script: vllm-container-deps.sh
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 2160
+  interval_seconds: 10
+
+resources:
+  gpu_type: "gb300"
+  gpus_per_node: 4
+  agg_nodes: 2
+  agg_workers: 1
+  gpus_per_agg: 8
+
+infra:
+  etcd_nats_dedicated_node: false
+  nats_max_payload_mb: 32
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+  args:
+    router-mode: "kv"
+    router-reset-states: true
+    router-temperature: 0.0
+    router-queue-threshold: 65536
+    active-decode-blocks-threshold: "None"
+    active-prefill-tokens-threshold: "None"
+    active-prefill-tokens-threshold-frac: "None"
+    tokenizer: "fastokens"
+
+backend:
+  type: vllm
+  connector: null
+  mooncake_kv_store:
+    store_config:
+      metadata_server: "P2PHANDSHAKE"
+      global_segment_size: "150GB"
+      local_buffer_size: "4GB"
+      protocol: "rdma"
+      device_name: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+      mode: "embedded"
+      enable_offload: false
+  aggregated_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "32"
+    TILELANG_CLEANUP_TEMP_FILES: "1"
+    VLLM_USE_NCCL_SYMM_MEM: "0"
+    TORCH_SYMMMEM: "NVSHMEM"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_USE_RUST_FRONTEND: "1"
+    VLLM_ALLREDUCE_USE_FLASHINFER: "1"
+    VLLM_FLASHINFER_ALLREDUCE_BACKEND: "auto"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_SPARSE_INDEXER_MAX_LOGITS_MB: "1024"
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: "1800"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    NCCL_P2P_LEVEL: "NVL"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-agg-mtp-{job_id}"
+  vllm_config:
+    aggregated:
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 8
+      pipeline-parallel-size: 1
+      disable-custom-all-reduce: true
+      enable-cumem-allocator: true
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      max-model-len: 1048576
+      max-num-seqs: 32
+      max-num-batched-tokens: 8192
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","mode":0}'
+      max-cudagraph-capture-size: 128
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      gpu-memory-utilization: 0.90
+      stream-interval: 10
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+
+sbatch_directives:
+  cpus-per-task: "144"
+  mem: "0"
+
+srun_options:
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: "/infmax-workspace"
+    RESULT_DIR: "/logs/agentic"
+    PORT: "8000"
+    # Keep aggregate workers in the multinode result schema so ingestion uses
+    # the zero decode-worker count instead of duplicating TP into P and D.
+    IS_MULTINODE: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
+    AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
+    HF_HUB_CACHE: "/hf_hub_cache"
+    WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic.yaml
new file mode 100644
index 0000000000..cc706a4aac
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic.yaml
@@ -0,0 +1,204 @@
+name: "svf-vllm-disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic"
+
+# Validated GB300 AgentX MTP3 low-latency topology: one DEP4 prefill worker
+# feeds one DEP8 decode worker at concurrency 128.
+
+model:
+  path: "deepseek-v4-pro"
+  container: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro"
+  container:
+    image: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  frameworks:
+    dynamo: "1.3.0.dev20260720"
+
+dynamo:
+  version: "1.3.0.dev20260720"
+  install: true
+
+setup_script: vllm-container-deps.sh
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 2160
+  interval_seconds: 10
+
+resources:
+  gpu_type: "gb300"
+  gpus_per_node: 4
+  het_jobs: false
+  spread_workers: false
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 4
+  gpus_per_decode: 8
+
+infra:
+  etcd_nats_dedicated_node: false
+  nats_max_payload_mb: 32
+
+environment:
+  # Mooncake prefix-block hashes must match across processes and nodes.
+  PYTHONHASHSEED: "0"
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+  args:
+    router-mode: "random"
+    router-session-affinity-ttl-secs: 900
+  env:
+    DYN_ROUTER_ACTIVE_REQUEST_EXPIRY_SECS: "3600"
+
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_node
+  mooncake_kv_store:
+    store_config:
+      metadata_server: "P2PHANDSHAKE"
+      global_segment_size: "150GB"
+      local_buffer_size: "4GB"
+      protocol: "rdma"
+      device_name: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+      mode: "embedded"
+      enable_offload: false
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector":"MultiConnector","kv_role":"kv_both","kv_connector_extra_config":{"connectors":[{"kv_connector":"NixlConnector","kv_role":"kv_both","kv_load_failure_policy":"fail","kv_buffer_device":"cuda","kv_connector_extra_config":{"enforce_handshake_compat":false,"enable_cross_layers_blocks":false}},{"kv_connector":"MooncakeStoreConnector","kv_role":"kv_both","kv_connector_extra_config":{"load_async":true,"lookup_async":true,"enable_cross_layers_blocks":false,"enable_offload":false}}]}}'
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13345
+      enable-cumem-allocator: true
+      enable-expert-parallel: true
+      enable-ep-weight-filter: true
+      max-model-len: 1048576
+      max-num-seqs: 64
+      max-num-batched-tokens: 8192
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      max-cudagraph-capture-size: 256
+      gpu-memory-utilization: 0.92
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      moe-backend: "deep_gemm_amxf4_mega_moe"
+      numa-bind: true
+      numa-bind-nodes: [0, 0, 1, 1]
+    decode:
+      kv-transfer-config: '{"kv_connector":"MultiConnector","kv_role":"kv_both","kv_connector_extra_config":{"connectors":[{"kv_connector":"NixlConnector","kv_role":"kv_both","kv_load_failure_policy":"fail","kv_buffer_device":"cuda","kv_connector_extra_config":{"enforce_handshake_compat":false,"enable_cross_layers_blocks":false}},{"kv_connector":"MooncakeStoreConnector","kv_role":"kv_consumer","kv_connector_extra_config":{"load_async":true,"lookup_async":false,"enable_lookup":false,"enable_cross_layers_blocks":false,"enable_offload":false}}]}}'
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-cumem-allocator: true
+      enable-expert-parallel: true
+      enable-ep-weight-filter: true
+      max-model-len: 1048576
+      max-num-seqs: 256
+      max-num-batched-tokens: 1024
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","mode":0}'
+      max-cudagraph-capture-size: 1024
+      gpu-memory-utilization: 0.92
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      moe-backend: "deep_gemm_amxf4_mega_moe"
+      numa-bind: true
+      numa-bind-nodes: [0, 0, 1, 1]
+  prefill_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "20"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_ALLREDUCE_USE_SYMM_MEM: "0"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    VLLM_RANDOMIZE_DP_DUMMY_INPUTS: "1"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_USE_BREAKABLE_CUDAGRAPH: "0"
+    VLLM_CONNECTOR_PREFETCH_DEPTH: "8"
+    VLLM_DSV4_MEGA_FP8_COMBINE: "1"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-1p1d-c128-mtp-{job_id}"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+  decode_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "20"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_ALLREDUCE_USE_SYMM_MEM: "0"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    VLLM_RANDOMIZE_DP_DUMMY_INPUTS: "1"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_DSV4_MEGA_FP8_COMBINE: "1"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-1p1d-c128-mtp-{job_id}"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+
+sbatch_directives:
+  cpus-per-task: "72"
+
+srun_options:
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: "/infmax-workspace"
+    RESULT_DIR: "/logs/agentic"
+    PORT: "8000"
+    IS_MULTINODE: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
+    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
+    AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
+    HF_HUB_CACHE: "/hf_hub_cache"
+    WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic.yaml
new file mode 100644
index 0000000000..fa856fe218
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic.yaml
@@ -0,0 +1,205 @@
+name: "svf-vllm-disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic"
+
+# Validated GB300 AgentX MTP3 efficiency topology: one DEP8 prefill worker and
+# one DEP8 decode worker at concurrency 384.
+
+model:
+  path: "deepseek-v4-pro"
+  container: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  precision: "fp4"
+
+identity:
+  model:
+    repo: "deepseek-ai/DeepSeek-V4-Pro"
+  container:
+    image: "vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f"
+  frameworks:
+    dynamo: "1.3.0.dev20260720"
+
+dynamo:
+  version: "1.3.0.dev20260720"
+  install: true
+
+setup_script: vllm-container-deps.sh
+
+slurm:
+  time_limit: "8:00:00"
+
+health_check:
+  max_attempts: 2160
+  interval_seconds: 10
+
+resources:
+  gpu_type: "gb300"
+  gpus_per_node: 4
+  het_jobs: false
+  spread_workers: false
+  prefill_nodes: 2
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 8
+  gpus_per_decode: 8
+
+infra:
+  etcd_nats_dedicated_node: false
+  nats_max_payload_mb: 32
+
+environment:
+  # Mooncake prefix-block hashes must match across processes and nodes.
+  PYTHONHASHSEED: "0"
+
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+  args:
+    router-mode: "random"
+    router-session-affinity-ttl-secs: 900
+  env:
+    DYN_ROUTER_ACTIVE_REQUEST_EXPIRY_SECS: "3600"
+
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_node
+  mooncake_kv_store:
+    store_config:
+      metadata_server: "P2PHANDSHAKE"
+      global_segment_size: "150GB"
+      local_buffer_size: "4GB"
+      protocol: "rdma"
+      device_name: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+      mode: "embedded"
+      enable_offload: false
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector":"MultiConnector","kv_role":"kv_both","kv_connector_extra_config":{"connectors":[{"kv_connector":"NixlConnector","kv_role":"kv_both","kv_load_failure_policy":"fail","kv_buffer_device":"cuda","kv_connector_extra_config":{"enforce_handshake_compat":false,"enable_cross_layers_blocks":false}},{"kv_connector":"MooncakeStoreConnector","kv_role":"kv_both","kv_connector_extra_config":{"load_async":true,"lookup_async":true,"enable_cross_layers_blocks":false,"enable_offload":false}}]}}'
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-cumem-allocator: true
+      enable-expert-parallel: true
+      enable-ep-weight-filter: true
+      max-model-len: 1048576
+      max-num-seqs: 32
+      max-num-batched-tokens: 8192
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      max-cudagraph-capture-size: 128
+      gpu-memory-utilization: 0.92
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      moe-backend: "deep_gemm_amxf4_mega_moe"
+      numa-bind: true
+      numa-bind-nodes: [0, 0, 1, 1]
+    decode:
+      kv-transfer-config: '{"kv_connector":"MultiConnector","kv_role":"kv_both","kv_connector_extra_config":{"connectors":[{"kv_connector":"NixlConnector","kv_role":"kv_both","kv_load_failure_policy":"fail","kv_buffer_device":"cuda","kv_connector_extra_config":{"enforce_handshake_compat":false,"enable_cross_layers_blocks":false}},{"kv_connector":"MooncakeStoreConnector","kv_role":"kv_consumer","kv_connector_extra_config":{"load_async":true,"lookup_async":false,"enable_lookup":false,"enable_cross_layers_blocks":false,"enable_offload":false}}]}}'
+      served-model-name: "deepseek-ai/DeepSeek-V4-Pro"
+      kv-cache-dtype: "fp8"
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-cumem-allocator: true
+      enable-expert-parallel: true
+      enable-ep-weight-filter: true
+      max-model-len: 1048576
+      max-num-seqs: 256
+      max-num-batched-tokens: 1024
+      trust-remote-code: true
+      no-enable-flashinfer-autotune: true
+      block-size: 256
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","mode":0}'
+      max-cudagraph-capture-size: 1024
+      gpu-memory-utilization: 0.92
+      no-disable-hybrid-kv-cache-manager: true
+      tokenizer-mode: "deepseek_v4"
+      attention-config: '{"backend":"FLASHINFER_MLA_SPARSE_DSV4","use_prefill_query_quantization":true,"use_fp4_indexer_cache":true}'
+      speculative-config: '{"method":"mtp","num_speculative_tokens":3,"rejection_sample_method":"synthetic","synthetic_acceptance_length":2.49}'
+      moe-backend: "deep_gemm_amxf4_mega_moe"
+      numa-bind: true
+      numa-bind-nodes: [0, 0, 1, 1]
+  prefill_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "20"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_ALLREDUCE_USE_SYMM_MEM: "0"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    VLLM_RANDOMIZE_DP_DUMMY_INPUTS: "1"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_USE_BREAKABLE_CUDAGRAPH: "0"
+    VLLM_CONNECTOR_PREFETCH_DEPTH: "8"
+    VLLM_DSV4_MEGA_FP8_COMBINE: "1"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-1p1d-mtp-{job_id}"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+  decode_environment:
+    HF_HUB_CACHE: "/hf_hub_cache"
+    HUGGINGFACE_HUB_CACHE: "/hf_hub_cache"
+    TRANSFORMERS_CACHE: "/hf_hub_cache"
+    VLLM_ENGINE_READY_TIMEOUT_S: "3600"
+    VLLM_RPC_TIMEOUT: "600000"
+    VLLM_LOG_STATS_INTERVAL: "1"
+    VLLM_V2_WARMUP_MAX_NUM_SEQS: "20"
+    VLLM_SERVER_DEV_MODE: "1"
+    VLLM_USE_V2_MODEL_RUNNER: "1"
+    VLLM_MOONCAKE_LOAD_RECV_THREADS: "20"
+    VLLM_ALLREDUCE_USE_SYMM_MEM: "0"
+    UCX_MEMTYPE_CACHE: "n"
+    UCX_NET_DEVICES: "mlx5_0:1,mlx5_1:1,mlx5_2:1,mlx5_3:1"
+    UCX_TLS: "rc,cuda_copy"
+    NCCL_CUMEM_ENABLE: "1"
+    NCCL_MNNVL_ENABLE: "1"
+    NCCL_NVLS_ENABLE: "1"
+    NCCL_IB_HCA: "mlx5_0,mlx5_1,mlx5_2,mlx5_3"
+    VLLM_RANDOMIZE_DP_DUMMY_INPUTS: "1"
+    VLLM_PREFIX_CACHE_RETENTION_INTERVAL: "32768"
+    VLLM_DSV4_MEGA_FP8_COMBINE: "1"
+    DG_JIT_CACHE_DIR: "/tmp/dg-cache-dsv4-gb300-1p1d-mtp-{job_id}"
+    MC_ENABLE_DEST_DEVICE_AFFINITY: "1"
+    MC_STORE_CLIENT_METRIC: "1"
+    MC_STORE_CLIENT_METRIC_INTERVAL: "5"
+    MC_TE_METRIC: "0"
+
+sbatch_directives:
+  cpus-per-task: "72"
+  mem: "0"
+
+srun_options:
+  container-remap-root: ""
+
+benchmark:
+  type: custom
+  command: bash /infmax-workspace/benchmarks/multi_node/agentic_srt.sh
+  env:
+    INFMAX_CONTAINER_WORKSPACE: "/infmax-workspace"
+    RESULT_DIR: "/logs/agentic"
+    PORT: "8000"
+    IS_MULTINODE: "true"
+    AIPERF_USE_DYNAMO_CONV_AWARE_ROUTING: "0"
+    AIPERF_HTTP_X_DYNAMO_SESSION_ID_FROM_CORRELATION_ID: "true"
+    AIPERF_AGENTIC_CACHE_WARMUP_DURATION: "600"
+    AIPERF_DATASET_MMAP_CACHE_DIR: "/aiperf_mmap_cache"
+    HF_HUB_CACHE: "/hf_hub_cache"
+    WEKA_LOADER_OVERRIDE: "semianalysis_cc_traces_weka_062126"
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index e7cf5b9e8e..36ae816809 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -7897,6 +7897,95 @@ dsv4-fp4-gb200-dynamo-vllm-agentic-2p1d-dep8-dep8:
           ep: 8
           dp-attn: true
 
+dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-agg:
+  image: vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f
+  model: deepseek-ai/DeepSeek-V4-Pro
+  model-prefix: dsv4
+  runner: cluster:gb300-nv
+  precision: fp4
+  framework: dynamo-vllm
+  router: { name: dynamo-router, version: "1.2.1" }
+  multinode: true
+  disagg: false
+  scenarios:
+    agentic-coding:
+    - search-space:
+      - spec-decoding: mtp
+        conc-list: [1]
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/deepseek-v4/agentic/agg-gb300-tp8-mtp-agentic.yaml"
+        # The aggregate worker also performs decode; keep the decode worker
+        # count at zero so result aggregation counts eight GPUs only once.
+        decode:
+          num-worker: 0
+          tp: 8
+          ep: 1
+          dp-attn: false
+      - spec-decoding: mtp
+        conc-list: [4]
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 1
+          dp-attn: false
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/deepseek-v4/agentic/agg-gb300-tp4-mtp-agentic.yaml"
+        # The aggregate worker also performs decode; keep the decode worker
+        # count at zero so result aggregation counts four GPUs only once.
+        decode:
+          num-worker: 0
+          tp: 4
+          ep: 1
+          dp-attn: false
+
+dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg:
+  image: vllm/vllm-openai:nightly-dev-arm64-cu13.0.1-426e59f
+  model: deepseek-ai/DeepSeek-V4-Pro
+  model-prefix: dsv4
+  runner: cluster:gb300-nv
+  precision: fp4
+  framework: dynamo-vllm
+  kv-p2p-transfer: nixl
+  multinode: true
+  disagg: true
+  scenarios:
+    agentic-coding:
+    - search-space:
+      - spec-decoding: mtp
+        conc-list: [128]
+        router: { name: dynamo-router, version: "1.3.0.dev20260720" }
+        prefill:
+          num-worker: 1
+          tp: 4
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep4-dep8-c128-mtp-agentic.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+      - spec-decoding: mtp
+        conc-list: [384]
+        router: { name: dynamo-router, version: "1.3.0.dev20260720" }
+        prefill:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/deepseek-v4/agentic/disagg-gb300-1p1d-dep8-dep8-c384-mtp-agentic.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 8
+          dp-attn: true
 dsv4-fp4-gb300-dynamo-sglang-agentic-agg:
   image: lmsysorg/sglang:nightly-dev-cu13-20260711-7de33ce8
   model: deepseek-ai/DeepSeek-V4-Pro
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 2c57c3c8a4..0759409403 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -5101,3 +5101,11 @@
     - "Recipe fix (model): adds mtp_flags (--speculative-algorithm EAGLE --speculative-eagle-topk 1) to the DeepSeek-V4-Pro-DI models.yaml entry; without it DECODE_MTP_SIZE>0 emitted --speculative-num-steps/--speculative-num-draft-tokens with no --speculative-algorithm, so sglang never engaged spec decoding. EAGLE with in-checkpoint draft (no draft-model-path); NEXTN (the V3/R1 keyword) crashes the dsv4 decode server at init"
     - "Recipe fix (launcher): build_server_config appended the speculative flags to the decode server only, so prefill never allocated the nextn (MTP draft) KV layer -- prefill registered 3 PD state components, decode 4. SGLang PD-disaggregation requires matching speculative config on both roles (the draft/MTP layers participate in both prefill KV computation and decode verification): v0.5.15 hard-fails the mismatch (mori 'state component count mismatch'), while v0.5.14 tolerated it silently but decode's nextn state was never seeded from prefill, corrupting greedy EAGLE verification (gsm8k 0.96 -> 0.88, worse at depth 2). Fix: apply the speculative flags to both prefill and decode"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2305/changes
+
+- config-keys:
+    - dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-agg
+    - dsv4-fp4-gb300-dynamo-vllm-agentic-mtp-disagg
+  description:
+    - "Add GB300 Dynamo-vLLM AgentX MTP3 points: aggregate TP8 c1, aggregate TP4 c4, P/D DEP4/DEP8 c128, and P/D DEP8/DEP8 c384; use upstream vLLM 426e59f, srt-slurm v1.0.36, and random session-affinity routing for P/D."
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2341
+  
diff --git a/runners/launch_gb300-nv.sh b/runners/launch_gb300-nv.sh
index d450cdb86d..e5a8d059b2 100644
--- a/runners/launch_gb300-nv.sh
+++ b/runners/launch_gb300-nv.sh
@@ -150,38 +150,16 @@ if [[ "$IS_AGENTIC" == "1" && $FRAMEWORK == "dynamo-sglang" && $MODEL_PREFIX ==
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/sglang/deepseek-v4/agentic" \
         recipes/sglang/deepseek-v4/agentic
 elif [[ "$IS_AGENTIC" == "1" ]]; then
-    # Agentic multi-node uses cquil11/srt-slurm-nv@cam/no-preflight-flag,
-    # a thin branch off NVIDIA/srt-slurm@127597c that adds one CLI flag
-    # (`srtctl apply --no-preflight`) — needed because:
-    #
-    #   - We want MODEL_PATH=/scratch/models/DeepSeek-V4-Pro (node-local
-    #     NVMe, fast) instead of the NFS path under /data/home/sa-shared.
-    #   - /scratch only exists on GB300 compute nodes; it is NOT mounted
-    #     on the GHA runner pod that invokes srtctl.
-    #   - srtctl's pre-submit model check (_preflight_model in
-    #     src/srtctl/core/validation.py) does a Path.is_dir() in-process
-    #     on the invoking node — so it fails before sbatch is ever
-    #     called with "Model alias 'X' resolved to '/scratch/...',
-    #     but that path is unavailable".
-    #   - --no-preflight skips just the optional Python-level FS check.
-    #     vLLM still fails loudly at runtime if the path is genuinely
-    #     missing on the compute node.
-    #
-    # All other upstream schema features we need are inherited from
-    # NVIDIA HEAD:
-    #   - BenchmarkType.CUSTOM + benchmark.command + benchmark.env
-    #     (hook that hands off to benchmarks/multi_node/agentic_srt.sh)
-    #   - DynamoConfig.wheel (so vllm recipes can pin the ai-dynamo wheel)
-    #   - sbatch_directives / srun_options (top-level recipe fields)
-    git clone https://github.com/cquil11/srt-slurm-nv.git "$SRT_REPO_DIR"
-    cd "$SRT_REPO_DIR"
-    # 854b3fd = --no-preflight flag
-    # 6e34b8b = benchmark_stage propagates srun_options (needed for
-    #           container-remap-root to reach the agentic_srt.sh srun)
-    git checkout 6e34b8b83229634d732e41a4e2d6595f46ef60b5
-    mkdir -p recipes/vllm/deepseek-v4/agentic
+    # Agentic recipes use NVIDIA/srt-slurm v1.0.36. This is the upstream
+    # version validated in InferenceX PR #2302 and includes per-node DP,
+    # matching Dynamo health counts, multi-node TP port handling, and
+    # Mooncake compatibility. Keep it pinned so sweeps are reproducible.
+    git clone --branch v1.0.36 --single-branch https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR" || exit 1
+    cd "$SRT_REPO_DIR" || exit 1
+
+    mkdir -p recipes/vllm/deepseek-v4/agentic || exit 1
     cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/deepseek-v4/agentic" \
-        recipes/vllm/deepseek-v4/agentic
+        recipes/vllm/deepseek-v4/agentic || exit 1
 elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "dsv4" ]]; then
     git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
     cd "$SRT_REPO_DIR"
diff --git a/utils/aiperf b/utils/aiperf
index 0d2aa0572a..ed2e040d0e 160000
--- a/utils/aiperf
+++ b/utils/aiperf
@@ -1 +1 @@
-Subproject commit 0d2aa0572ac685943d38c580675c4a61023581d3
+Subproject commit ed2e040d0ed0976bbf632e5095d0b1ccebf8f6a2