diff --git a/scripts/train.py b/scripts/train.py
index 0697baaef..8b5d4bfc8 100644
--- a/scripts/train.py
+++ b/scripts/train.py
@@ -620,10 +620,9 @@ def parse_args():
     parser.add_argument(
         "--draft-arch",
         type=str,
-        default="llama",
+        default="qwen3",
         choices=list(DRAFT_ARCH_CONFIGS.keys()),
-        help="Architecture for draft decoder layers. Defaults to 'llama'. "
-        "Note: only 'llama' is currently supported in vLLM for inference.",
+        help="Architecture for draft decoder layers. Defaults to 'qwen3'.",
     )
     parser.add_argument(
         "--draft-hidden-act",
diff --git a/src/speculators/models/eagle3/config.py b/src/speculators/models/eagle3/config.py
index a3201c452..dbcdb7bdb 100644
--- a/src/speculators/models/eagle3/config.py
+++ b/src/speculators/models/eagle3/config.py
@@ -2,7 +2,7 @@
 
 from pydantic import Field, field_serializer, field_validator
 from transformers import AutoConfig, PretrainedConfig
-from transformers.models.llama.configuration_llama import LlamaConfig
+from transformers.models.qwen3.configuration_qwen3 import Qwen3Config
 
 from speculators import SpeculatorModelConfig
 
@@ -31,7 +31,7 @@ class Eagle3SpeculatorConfig(SpeculatorModelConfig):
     )
 
     transformer_layer_config: PretrainedConfig = Field(
-        default_factory=LlamaConfig,
+        default_factory=Qwen3Config,
         description="Configuration for the transformer decoder layer",
     )
 
@@ -79,7 +79,7 @@ def serialize_transformer_config(self, value: PretrainedConfig) -> dict:
     def validate_transformer_config(cls, value: Any) -> PretrainedConfig:
         """Validate and convert transformer config."""
         if isinstance(value, dict):
-            config_class: type[PretrainedConfig] = LlamaConfig
+            config_class: type[PretrainedConfig] = Qwen3Config
             if "model_type" in value:
                 config_class = AutoConfig.for_model(
                     model_type=value["model_type"]