diff --git a/vllm/entrypoints/generate/generative_scoring/api_router.py b/vllm/entrypoints/generate/generative_scoring/api_router.py
index d3e2316fcd3c..70fbea539d37 100644
--- a/vllm/entrypoints/generate/generative_scoring/api_router.py
+++ b/vllm/entrypoints/generate/generative_scoring/api_router.py
@@ -6,6 +6,7 @@
 from fastapi.responses import JSONResponse
 
 from vllm.entrypoints.generate.generative_scoring.serving import (
+    GenerativeScoringRequest,
     GenerativeScoringResponse,
     ServingGenerativeScoring,
 )
@@ -36,21 +37,16 @@ def generative_scoring(request: Request) -> ServingGenerativeScoring | None:
 )
 @with_cancellation
 @load_aware_call
-async def create_generative_scoring(raw_request: Request):
+async def create_generative_scoring(
+    request: GenerativeScoringRequest, raw_request: Request
+):
     handler = generative_scoring(raw_request)
     if handler is None:
         raise NotImplementedError(
             "The model does not support the Generative Scoring API"
         )
 
-    raw_body = await raw_request.json()
-
-    from vllm.entrypoints.generate.generative_scoring.serving import (
-        GenerativeScoringRequest,
-    )
-
-    gen_request = GenerativeScoringRequest(**raw_body)
-    result = await handler.create_generative_scoring(gen_request, raw_request)
+    result = await handler.create_generative_scoring(request, raw_request)
 
     if isinstance(result, ErrorResponse):
         return JSONResponse(content=result.model_dump(), status_code=result.error.code)