diff --git a/python/sglang/srt/speculative/draft_worker_common.py b/python/sglang/srt/speculative/draft_worker_common.py
index 28a150f7375f..058f0f7824ed 100644
--- a/python/sglang/srt/speculative/draft_worker_common.py
+++ b/python/sglang/srt/speculative/draft_worker_common.py
@@ -71,6 +71,7 @@ def build_draft_tp_worker(
     target_model_config: ModelConfig,
     algo_label: str,
     attention_backend_override: Optional[str] = None,
+    draft_worker_cls: type[TpModelWorker] = TpModelWorker,
 ) -> DraftWorkerBundle:
     # An override names a draft-specific backend the caller has already
     # validated (e.g. a self-drafting architecture); it skips the generic
@@ -88,7 +89,7 @@ def build_draft_tp_worker(
     # workers run the draft outside speculative_moe_backend_context, so a
     # construction-only swap would build and execute under different backends.
     with draft_model_build_scope():
-        draft_worker = TpModelWorker(
+        draft_worker = draft_worker_cls(
             server_args=server_args,
             gpu_id=gpu_id,
             ps=ps,
diff --git a/python/sglang/srt/speculative/dspark_components/dspark_worker_v2.py b/python/sglang/srt/speculative/dspark_components/dspark_worker_v2.py
index 666cfd8c0b1d..d3024c4454b1 100644
--- a/python/sglang/srt/speculative/dspark_components/dspark_worker_v2.py
+++ b/python/sglang/srt/speculative/dspark_components/dspark_worker_v2.py
@@ -82,6 +82,7 @@ def __init__(
         ps: ParallelState,
         nccl_port: int,
         target_worker: TpModelWorker,
+        draft_worker_cls: type[TpModelWorker] = TpModelWorker,
     ):
         super().__init__()
 
@@ -124,6 +125,7 @@ def __init__(
                 attention_backend_override=(
                     DSV4_DRAFT_ATTENTION_BACKEND if self._draft_is_moe else None
                 ),
+                draft_worker_cls=draft_worker_cls,
             )
         self._draft_worker = bundle.draft_worker
         self.draft_model_runner = bundle.draft_model_runner