diff --git a/vllm_omni/entrypoints/omni_base.py b/vllm_omni/entrypoints/omni_base.py
index ab2bee01ba1..0d1a8f4089e 100644
--- a/vllm_omni/entrypoints/omni_base.py
+++ b/vllm_omni/entrypoints/omni_base.py
@@ -177,7 +177,12 @@ def __init__(
         # override the deploy YAML's ``async_chunk: true`` default.
         async_chunk = kwargs.get("async_chunk")
         output_modalities = kwargs.pop("output_modalities", None)
-        diffusion_batch_size: int = kwargs.pop("diffusion_batch_size", 1)
+        # Stage init overwrites ``od_config.max_num_seqs`` with this value.
+        # ``vllm serve --omni --max-num-seqs N`` is the documented request-batch
+        # knob; use it when ``diffusion_batch_size`` is not passed explicitly.
+        explicit_batch = kwargs.pop("diffusion_batch_size", None)
+        max_num_seqs = kwargs.get("max_num_seqs") or 1
+        diffusion_batch_size = int(explicit_batch if explicit_batch is not None else max_num_seqs)
 
         if "log_requests" in kwargs:
             raise TypeError("`log_requests` has been removed in Omni/AsyncOmni. Use `log_stats`.")