diff --git a/src/art/megatron/model_support/registry.py b/src/art/megatron/model_support/registry.py
index 3bd93f34e..855b034cc 100644
--- a/src/art/megatron/model_support/registry.py
+++ b/src/art/megatron/model_support/registry.py
@@ -151,6 +151,7 @@
         "Qwen/Qwen3.5-9B",
         "Qwen/Qwen3.5-27B",
         "Qwen/Qwen3.6-27B",
+        "Qwen/Qwen3.8-27B",
     ),
     default_target_modules=_QWEN3_5_DENSE_TARGET_MODULES,
     native_vllm_lora_status=_VALIDATED_NATIVE_VLLM_LORA_STATUS,
diff --git a/src/art/trainer_rank/_impl.py b/src/art/trainer_rank/_impl.py
index fa701a4f2..9249f55d3 100644
--- a/src/art/trainer_rank/_impl.py
+++ b/src/art/trainer_rank/_impl.py
@@ -1377,7 +1377,7 @@ def _validate_checkpoint_adapter_config(
             raise TypeError(
                 "adapter_config['base_model_name_or_path'] must be a string"
             )
-        if base_model.startswith(("Qwen/Qwen3.5-", "Qwen/Qwen3.6-")):
+        if base_model.startswith(("Qwen/Qwen3.5-", "Qwen/Qwen3.6-", "Qwen/Qwen3.8-")):
             dimensions = {
                 "num_attention_heads": getattr(
                     self.runtime.provider, "num_attention_heads", None
diff --git a/tests/integration/megatron/model_support/test_provider_support.py b/tests/integration/megatron/model_support/test_provider_support.py
index 4a5b0abfd..f3c7e66e8 100644
--- a/tests/integration/megatron/model_support/test_provider_support.py
+++ b/tests/integration/megatron/model_support/test_provider_support.py
@@ -146,6 +146,16 @@ def test_qwen35_9b_uses_qwen35_dense_support() -> None:
     assert handler.key == "qwen3_5_dense"
 
 
+def test_qwen38_27b_uses_qwen35_dense_support() -> None:
+    spec = get_model_support_spec("Qwen/Qwen3.8-27B")
+    handler = get_model_support_handler("Qwen/Qwen3.8-27B")
+
+    assert spec.key == "qwen3_5_dense"
+    assert spec.is_moe is False
+    assert spec.native_vllm_lora_status == "validated"
+    assert handler.key == "qwen3_5_dense"
+
+
 def test_meta_llama_32_1b_instruct_uses_llama3_dense_support() -> None:
     model = "meta-llama/Llama-3.2-1B-Instruct"
     spec = get_model_support_spec(model)
diff --git a/tests/unit/test_trainer_rank_validation.py b/tests/unit/test_trainer_rank_validation.py
index f1a78a20b..2c4fae3cf 100644
--- a/tests/unit/test_trainer_rank_validation.py
+++ b/tests/unit/test_trainer_rank_validation.py
@@ -968,7 +968,13 @@ def test_checkpoint_slot_adapter_config_is_validated_and_copied() -> None:
         trainer._validate_checkpoint_adapter_config("student", {"r": 8}, alpha=None)
 
 
-def test_qwen35_checkpoint_adapter_config_captures_attention_dimensions() -> None:
+@pytest.mark.parametrize(
+    "base_model",
+    ("Qwen/Qwen3.5-4B", "Qwen/Qwen3.6-27B", "Qwen/Qwen3.8-27B"),
+)
+def test_qwen35_checkpoint_adapter_config_captures_attention_dimensions(
+    base_model: str,
+) -> None:
     runtime = _runtime()
     runtime.provider.num_attention_heads = 16
     runtime.provider.num_query_groups = 4
@@ -978,7 +984,7 @@ def test_qwen35_checkpoint_adapter_config_captures_attention_dimensions() -> Non
     retained = trainer._validate_checkpoint_adapter_config(
         "student",
         {
-            "base_model_name_or_path": "Qwen/Qwen3.5-4B",
+            "base_model_name_or_path": base_model,
             "r": 8,
             "lora_alpha": 16,
             "target_modules": ["q_proj"],