diff --git a/vllm/v1/worker/gpu/model_states/__init__.py b/vllm/v1/worker/gpu/model_states/__init__.py
index 06b5a92c395..b096fcaf5e6 100644
--- a/vllm/v1/worker/gpu/model_states/__init__.py
+++ b/vllm/v1/worker/gpu/model_states/__init__.py
@@ -13,7 +13,10 @@ def init_model_state(
     encoder_cache: EncoderCache | None,
     device: torch.device,
 ):
-    if "WhisperForConditionalGeneration" in vllm_config.model_config.architectures:
+    if (
+        "WhisperForConditionalGeneration" in vllm_config.model_config.architectures
+        or "CohereAsrForConditionalGeneration" in vllm_config.model_config.architectures
+    ):
         from vllm.v1.worker.gpu.model_states.whisper import WhisperModelState
 
         return WhisperModelState(vllm_config, model, encoder_cache, device)