diff --git a/studio/backend/core/inference/inference.py b/studio/backend/core/inference/inference.py
index d194f08045..780a399637 100644
--- a/studio/backend/core/inference/inference.py
+++ b/studio/backend/core/inference/inference.py
@@ -214,8 +214,9 @@ class InferenceBackend:
                     self.models[model_name]["model"] = model
                     self.models[model_name]["tokenizer"] = tokenizer
 
-                # Load the external codec for TTS audio types (Whisper is ASR, no codec needed)
-                if audio_type != "whisper":
+                # Load the external codec for TTS audio types
+                # (Whisper is ASR, audio_vlm is audio input — neither needs a codec)
+                if audio_type not in ("whisper", "audio_vlm"):
                     model_repo_path = self.models[model_name].get("model_repo_path")
                     self._audio_codec_manager.load_codec(audio_type, self.device, model_repo_path=model_repo_path)
 
@@ -835,6 +836,7 @@ class InferenceBackend:
             tokenize=True,
             return_dict=True,
             return_tensors="pt",
+            truncation=False,
         ).to(self.device)
 
         try:
diff --git a/studio/backend/utils/models/model_config.py b/studio/backend/utils/models/model_config.py
index 05817e8ccd..f301e42a78 100644
--- a/studio/backend/utils/models/model_config.py
+++ b/studio/backend/utils/models/model_config.py
@@ -1258,7 +1258,7 @@ class ModelConfig:
                 is_cached=True,  # Local LoRAs are always "cached"
                 is_vision=is_vision,
                 is_lora=True,
-                is_audio=audio_type is not None,
+                is_audio=audio_type is not None and audio_type != 'audio_vlm',
                 audio_type=audio_type,
                 has_audio_input=is_audio_input_type(audio_type),
                 base_model=base_model,
@@ -1456,7 +1456,7 @@ class ModelConfig:
             is_cached=is_model_cached(identifier) if not is_local else True,
             is_vision=vision,
             is_lora=is_lora,
-            is_audio=audio_type_val is not None,
+            is_audio=audio_type_val is not None and audio_type_val != 'audio_vlm',
             audio_type=audio_type_val,
             has_audio_input=has_audio_in,
             base_model=base_model,