diff --git a/sdk/rt/speechmatics/rt/_models.py b/sdk/rt/speechmatics/rt/_models.py
index cf2540e..c035834 100644
--- a/sdk/rt/speechmatics/rt/_models.py
+++ b/sdk/rt/speechmatics/rt/_models.py
@@ -41,6 +41,7 @@ class OperatingPoint(str, Enum):
     ENHANCED = "enhanced"
     STANDARD = "standard"
     MELIA_1 = "melia-1"
+    OAK_1 = "oak-1"
 
 
 class Model(str, Enum):
@@ -49,6 +50,7 @@ class Model(str, Enum):
     ENHANCED = "enhanced"
     STANDARD = "standard"
     MELIA_1 = "melia-1"
+    OAK_1 = "oak-1"
 
 
 @dataclass
@@ -399,7 +401,7 @@ class TranscriptionConfig:
             Recognition is restricted to these languages, which improves accuracy on multilingual
             audio. Typically used with ``language="multi"``. Hints take precedence over ``language``,
             so the transcript may not contain the language named there.
-            Applicable only for the next-gen models (e.g. melia-1).
+            Applicable only for the next-gen models (e.g. melia-1, oak-1).
             Defaults to None.
         language_hints_strict: (Optional) Controls how strictly ``language_hints`` is applied, and is
             only valid alongside it.
diff --git a/tests/rt/test_models.py b/tests/rt/test_models.py
index e4cd974..69f5a98 100644
--- a/tests/rt/test_models.py
+++ b/tests/rt/test_models.py
@@ -1,6 +1,15 @@
+from speechmatics.rt._models import Model
 from speechmatics.rt._models import TranscriptionConfig
 
 
+class TestModelToDict:
+    def test_oak_model_gets_serialized(self):
+        config = TranscriptionConfig(model=Model.OAK_1)
+        result = config.to_dict()
+        assert result["model"] == Model.OAK_1
+        assert "operating_point" not in result
+
+
 class TestLanguageHintsToDict:
     def test_language_hints_serializes_correctly(self):
         config = TranscriptionConfig(language_hints=["en", "fr"])