diff --git a/nemo_deploy/llm/megatronllm_deployable.py b/nemo_deploy/llm/megatronllm_deployable.py
index 8cd59d50e1..d4359d1e26 100755
--- a/nemo_deploy/llm/megatronllm_deployable.py
+++ b/nemo_deploy/llm/megatronllm_deployable.py
@@ -22,7 +22,7 @@
 import torch.distributed
 from jinja2 import Template
 from megatron.core.inference.common_inference_params import CommonInferenceParams
-from megatron.core.inference.inference_request import InferenceRequest
+from megatron.core.inference.inference_request import DynamicInferenceRequestRecord, InferenceRequest
 
 from nemo_deploy import ITritonDeployable
 from nemo_deploy.llm.inference.inference_base import create_mcore_engine
@@ -470,6 +470,10 @@ def _infer_fn(
         )
 
         results = self.generate(prompts, inference_params)
+        # Handle DynamicInferenceRequestRecord objects by merging them into a single request
+        results = [
+            r.merge(self.mcore_tokenizer) if isinstance(r, DynamicInferenceRequestRecord) else r for r in results
+        ]
         if echo:
             output_texts = [r.prompt + r.generated_text if text_only else r for r in results]
         else: