diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep4-dep8-c1024.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep4-dep8-c1024.yaml
new file mode 100644
index 0000000000..5cb6c798b6
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep4-dep8-c1024.yaml
@@ -0,0 +1,112 @@
+name: kimi-k2.6-vllm-disagg-b200-1p1d-dep4-dep8-c1024
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 4
+  gpus_per_decode: 8
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep4"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.94
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.9
+      stream-interval: 50
+      max-cudagraph-capture-size: 1024
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '1024'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-dep8-c2048.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-dep8-c2048.yaml
new file mode 100644
index 0000000000..e3c08dc477
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-dep8-c2048.yaml
@@ -0,0 +1,112 @@
+name: kimi-k2.6-vllm-disagg-b200-1p1d-dep8-dep8-c2048
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 8
+  gpus_per_decode: 8
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep8"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 16384
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.9
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.9
+      stream-interval: 50
+      max-cudagraph-capture-size: 1024
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '2048'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-tp8-c1.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-tp8-c1.yaml
new file mode 100644
index 0000000000..1c0abe2fa0
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-tp8-c1.yaml
@@ -0,0 +1,110 @@
+name: kimi-k2.6-vllm-disagg-b200-1p1d-dep8-tp8-c1
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 1
+  prefill_workers: 1
+  decode_workers: 1
+  gpus_per_prefill: 8
+  gpus_per_decode: 8
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '0'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep8"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 16384
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.9
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 8
+      pipeline-parallel-size: 1
+      disable-custom-all-reduce: true
+      max-model-len: 9216
+      max-num-seqs: 1024
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.9
+      stream-interval: 50
+      max-cudagraph-capture-size: 256
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '1'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p4d-dep4-tp4-c512.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p4d-dep4-tp4-c512.yaml
new file mode 100644
index 0000000000..1ca87ad9b8
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p4d-dep4-tp4-c512.yaml
@@ -0,0 +1,109 @@
+name: kimi-k2.6-vllm-disagg-b200-1p4d-dep4-tp4-c512
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 2
+  prefill_workers: 1
+  decode_workers: 4
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep4"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.94
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      max-model-len: 9216
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.93
+      stream-interval: 50
+      max-cudagraph-capture-size: 2048
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '512'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c128.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c128.yaml
new file mode 100644
index 0000000000..44f8ff7849
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c128.yaml
@@ -0,0 +1,109 @@
+name: kimi-k2.6-vllm-disagg-b200-1p8d-dep4-tp4-c128
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 4
+  prefill_workers: 1
+  decode_workers: 8
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep4"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.94
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      max-model-len: 9216
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.93
+      stream-interval: 50
+      max-cudagraph-capture-size: 2048
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '128'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c32.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c32.yaml
new file mode 100644
index 0000000000..c748107fec
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c32.yaml
@@ -0,0 +1,109 @@
+name: kimi-k2.6-vllm-disagg-b200-1p8d-dep4-tp4-c32
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 1
+  decode_nodes: 4
+  prefill_workers: 1
+  decode_workers: 8
+  gpus_per_prefill: 4
+  gpus_per_decode: 4
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both", "engine_id": "kimi-k26-prefill-dep4"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 4
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.93
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 4
+      pipeline-parallel-size: 1
+      max-model-len: 9216
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.93
+      stream-interval: 50
+      max-cudagraph-capture-size: 2048
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '32'
+  req_rate: inf
+  use_chat_template: true
diff --git a/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-2p1d-dep8-dep8-c8192.yaml b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-2p1d-dep8-dep8-c8192.yaml
new file mode 100644
index 0000000000..54e623d767
--- /dev/null
+++ b/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-2p1d-dep8-dep8-c8192.yaml
@@ -0,0 +1,112 @@
+name: kimi-k2.6-vllm-disagg-b200-2p1d-dep8-dep8-c8192
+model:
+  path: kimi-k2.6-nvfp4
+  container: vllm/vllm-openai:v0.25.1
+  precision: fp4
+dynamo:
+  wheel: 1.3.0.dev20260721
+  install: true
+resources:
+  gpu_type: b200
+  gpus_per_node: 8
+  prefill_nodes: 2
+  decode_nodes: 1
+  prefill_workers: 2
+  decode_workers: 1
+  gpus_per_prefill: 8
+  gpus_per_decode: 8
+infra:
+  etcd_nats_dedicated_node: true
+frontend:
+  type: dynamo
+  enable_multiple_frontends: false
+backend:
+  type: vllm
+  connector: null
+  dp_launch_mode: per_gpu
+  prefill_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: '900'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  decode_environment:
+    PYTHONUNBUFFERED: '1'
+    VLLM_MEMORY_PROFILER_ESTIMATE_CUDAGRAPHS: '0'
+    VLLM_USE_FLASHINFER_MOE_FP4: '1'
+    VLLM_USE_NCCL_SYMM_MEM: '0'
+    NCCL_CUMEM_ENABLE: '1'
+    NCCL_WATCHDOG_TIMEOUT: '1800'
+    TORCH_NCCL_HEARTBEAT_TIMEOUT_SEC: '1800'
+    TORCH_CUDA_ARCH_LIST: '10.0'
+    PYTHONNOUSERSITE: '1'
+    UCX_MEMTYPE_CACHE: n
+    UCX_MEMTYPE_REG_WHOLE: n
+    UCX_CUDA_IPC_ENABLE_MNNVL: n
+    UCX_MAX_RMA_RAILS: '1'
+    UCX_MAX_RNDV_RAILS: '1'
+    UCX_RNDV_SCHEME: put_zcopy
+  vllm_config:
+    prefill:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13346
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      enforce-eager: true
+      compilation-config: '{"custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      max-num-batched-tokens: 16384
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      attention-config: '{"mla_prefill_backend": "TRTLLM_RAGGED"}'
+      gpu-memory-utilization: 0.9
+    decode:
+      kv-transfer-config: '{"kv_connector": "NixlConnector", "kv_role": "kv_both"}'
+      served-model-name: nvidia/Kimi-K2.6-NVFP4
+      kv-cache-dtype: fp8
+      tensor-parallel-size: 1
+      pipeline-parallel-size: 1
+      data-parallel-size: 8
+      data-parallel-rpc-port: 13345
+      enable-expert-parallel: true
+      max-model-len: 10240
+      max-num-seqs: 2048
+      max-num-batched-tokens: 8192
+      safetensors-load-strategy: prefetch
+      trust-remote-code: true
+      no-enable-prefix-caching: true
+      no-enable-flashinfer-autotune: true
+      async-scheduling: true
+      attention-backend: FLASHINFER_MLA
+      block-size: 128
+      compilation-config: '{"cudagraph_mode":"FULL_DECODE_ONLY","custom_ops":["+quant_fp8","+rms_norm","+rotary_embedding"],"pass_config":{"fuse_attn_quant":true,"fuse_allreduce_rms":true}}'
+      gpu-memory-utilization: 0.9
+      stream-interval: 50
+      max-cudagraph-capture-size: 1024
+benchmark:
+  type: sa-bench
+  isl: 8192
+  osl: 1024
+  concurrencies: '8192'
+  req_rate: inf
+  use_chat_template: true
diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml
index 9fc14fd4e8..a30af59641 100644
--- a/configs/nvidia-master.yaml
+++ b/configs/nvidia-master.yaml
@@ -4977,6 +4977,114 @@ kimik2.5-fp4-gb200-dynamo-vllm:
           ep: 16
           dp-attn: true
 
+kimik2.6-fp4-b200-dynamo-vllm:
+  image: vllm/vllm-openai:v0.25.1
+  model: nvidia/Kimi-K2.6-NVFP4
+  model-prefix: kimik2.6
+  runner: b200-multinode
+  precision: fp4
+  framework: dynamo-vllm
+  router: { name: dynamo-router, version: "1.3.0.dev20260721" }
+  kv-p2p-transfer: nixl
+  multinode: true
+  disagg: true
+  scenarios:
+    fixed-seq-len:
+    - isl: 8192
+      osl: 1024
+      search-space:
+      - conc-list: [1]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-tp8-c1.yaml"
+        decode:
+          num-worker: 1
+          tp: 8
+          ep: 1
+          dp-attn: false
+      - conc-list: [32]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c32.yaml"
+        decode:
+          num-worker: 8
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [128]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p8d-dep4-tp4-c128.yaml"
+        decode:
+          num-worker: 8
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [512]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p4d-dep4-tp4-c512.yaml"
+        decode:
+          num-worker: 4
+          tp: 4
+          ep: 1
+          dp-attn: false
+      - conc-list: [1024]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 4
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep4-dep8-c1024.yaml"
+        decode:
+          num-worker: 1
+          tp: 1
+          ep: 8
+          dp-attn: true
+      - conc-list: [2048]
+        prefill:
+          num-worker: 1
+          tp: 1
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-1p1d-dep8-dep8-c2048.yaml"
+        decode:
+          num-worker: 1
+          tp: 1
+          ep: 8
+          dp-attn: true
+      - conc-list: [8192]
+        prefill:
+          num-worker: 2
+          tp: 1
+          ep: 8
+          dp-attn: true
+          additional-settings:
+          - "CONFIG_FILE=recipes/vllm/kimi-k2.6/b200-fp4/8k1k/disagg-b200-2p1d-dep8-dep8-c8192.yaml"
+        decode:
+          num-worker: 1
+          tp: 1
+          ep: 8
+          dp-attn: true
+
 dsv4-fp4-b200-dynamo-vllm:
   image: vllm/vllm-openai:v0.23.0
   model: deepseek-ai/DeepSeek-V4-Pro
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 8f5831507e..4c22ed9198 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -5339,3 +5339,11 @@
     - "Pass --use-chat-template to the benchmark, as required for EAGLE-style speculative decoding"
     - "Validated on the node through the real launcher, every request successful: TEP4 concurrency 64 640/640 at 6895 tok/s (11% above the non-MTP arm's 6198 tok/s, mean TTFT 27.2s to 9.6s), and TEP4 concurrency 1 10/10 at 1233 tok/s (83% above the non-MTP arm's 672 tok/s, mean TPOT 12.6ms to 6.5ms)"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2312
+
+- config-keys:
+    - kimik2.6-fp4-b200-dynamo-vllm
+  description:
+    - "Add Kimi K2.6 NVFP4 B200 Dynamo vLLM disaggregated configuration."
+    - "Add seven checked-in srt-slurm recipes selected one-to-one by the 8K/1K prefill/decode topology and concurrency entries."
+    - "Image: vllm/vllm-openai:v0.25.1"
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2360
diff --git a/runners/launch_b200-dgxc.sh b/runners/launch_b200-dgxc.sh
index b304ad4c1f..be68ce073b 100644
--- a/runners/launch_b200-dgxc.sh
+++ b/runners/launch_b200-dgxc.sh
@@ -6,13 +6,11 @@ SLURM_ACCOUNT="benchmark"
 
 set -x
 
-# MODEL_PATH: Override with pre-downloaded paths on the shared Lustre tree.
+# MODEL_PATH: Override with pre-downloaded paths on cluster-accessible storage.
 # Bench scripts and srt-slurm yaml configs specify HuggingFace model IDs for
-# portability, but we resolve to /lustre/fsw/models/* here to avoid repeated
+# portability, but we resolve to pre-staged paths here to avoid repeated
 # downloading on every dgxc node. Runs for both single-node and multinode
 # launches.
-# NOTE: per-node /raid/models/* would be faster but is only populated on a
-# subset of dgxc nodes today, so we use Lustre for reliability.
 if [[ $MODEL_PREFIX == "dsr1" && $PRECISION == "fp4" ]]; then
     export MODEL_PATH="/scratch/fsw/models/DeepSeek-R1-0528-NVFP4-v2"
     export SRT_SLURM_MODEL_PREFIX="dsr1"
@@ -54,6 +52,9 @@ elif [[ $MODEL_PREFIX == "kimik2.5" && $PRECISION == "int4" ]]; then
 elif [[ $MODEL_PREFIX == "kimik2.5" && $PRECISION == "fp4" ]]; then
     export MODEL_PATH="/lustre/fsw/models/Kimi-K2.5-NVFP4"
     export SRT_SLURM_MODEL_PREFIX="kimik2.5-fp4"
+elif [[ $MODEL_PREFIX == "kimik2.6" && $PRECISION == "fp4" ]]; then
+    export MODEL_PATH="/lustre/fsw/models/Kimi-K2.6-NVFP4"
+    export SRT_SLURM_MODEL_PREFIX="kimi-k2.6-nvfp4"
 elif [[ $MODEL_PREFIX == "minimaxm2.5" && $PRECISION == "fp8" ]]; then
     export MODEL_PATH="/lustre/fsw/models/MiniMax-M2.5"
     export SRT_SLURM_MODEL_PREFIX="minimax-m2.5-fp8"
@@ -127,6 +128,12 @@ if [[ "$IS_MULTINODE" == "true" ]]; 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 == "kimik2.6" && $PRECISION == "fp4" ]]; then
+        git clone --branch main --single-branch https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
+        cd "$SRT_REPO_DIR" || exit 1
+        git checkout c180328b98c3793ca84a1e24a030f90545eb7d5d || exit 1
+        mkdir -p recipes/vllm/kimi-k2.6
+        cp -rT "$GITHUB_WORKSPACE/benchmarks/multi_node/srt-slurm-recipes/vllm/kimi-k2.6" recipes/vllm/kimi-k2.6
     elif [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "minimaxm3" && $PRECISION == "fp4" ]]; then
         git clone https://github.com/NVIDIA/srt-slurm.git "$SRT_REPO_DIR"
         cd "$SRT_REPO_DIR" || exit 1
@@ -299,7 +306,14 @@ EOF
     # so large-model loads (e.g. DSR1-FP8 ~680GB off shared FS) finish in time.
     # Uses ${CONFIG_FILE%%:*} because CONFIG_FILE may carry an :override[N] suffix.
     sed -i 's/^  max_attempts: [0-9]*/  max_attempts: 720/' "${CONFIG_FILE%%:*}"
-    SRTCTL_OUTPUT=$(srtctl apply -f "$CONFIG_FILE" --tags "b200,${MODEL_PREFIX},${PRECISION},${ISL}x${OSL},infmax-$(date +%Y%m%d)" 2>&1)
+
+    SRTCTL_PREFLIGHT_ARGS=()
+    # Kimi K2.6 weights are staged on the Slurm compute nodes, not the login node.
+    if [[ $FRAMEWORK == "dynamo-vllm" && $MODEL_PREFIX == "kimik2.6" && $PRECISION == "fp4" ]]; then
+        SRTCTL_PREFLIGHT_ARGS+=(--no-preflight)
+    fi
+
+    SRTCTL_OUTPUT=$(srtctl apply -f "$CONFIG_FILE" "${SRTCTL_PREFLIGHT_ARGS[@]}" --tags "b200,${MODEL_PREFIX},${PRECISION},${ISL}x${OSL},infmax-$(date +%Y%m%d)" 2>&1)
     echo "$SRTCTL_OUTPUT"
 
     # Extract JOB_ID from srtctl output