diff --git a/docs/web/integrations/embedding/openai.mdx b/docs/web/integrations/embedding/openai.mdx
index fbfa62d9d..4dd1ee8ba 100644
--- a/docs/web/integrations/embedding/openai.mdx
+++ b/docs/web/integrations/embedding/openai.mdx
@@ -10,7 +10,7 @@ LanceDB registers the OpenAI embeddings function in the registry by default, as
 | Parameter | Type | Default Value | Description |
 |---|---|---|---|
 | `name` | `str` | `"text-embedding-ada-002"` | The name of the model. |
-| `dim` | `int` |  Model default   | Requested output dimensions for models that support shortening embeddings. Use `func.ndims()` when defining the vector field. |
+| `dim` | `int` |  Model default   | Requested output dimensions for models that support shortening embeddings. Required for model names other than the OpenAI ones listed by `model_names()` (for example a model served behind an OpenAI-compatible `base_url`), where it is the model's embedding dimension. Use `func.ndims()` when defining the vector field. |
 | `use_azure` | bool | `False` | Set true to use Azure OpenAI SDK |
 
 Set your OpenAI API key before running the example:
diff --git a/python/python/lancedb/embeddings/openai.py b/python/python/lancedb/embeddings/openai.py
index d4ecbdf3f..8592598cc 100644
--- a/python/python/lancedb/embeddings/openai.py
+++ b/python/python/lancedb/embeddings/openai.py
@@ -36,6 +36,7 @@ class OpenAIEmbeddings(TextEmbeddingFunction):
     embedding_function = openai.create(
         name="<ollama-embedding-model-name>",
         base_url="http://localhost:11434",
+        dim=768,  # the model's embedding dimension; required for unlisted names
         )
     ```
 
@@ -77,8 +78,15 @@ class OpenAIEmbeddings(TextEmbeddingFunction):
             return self.dim or 3072
         elif self.name == "text-embedding-3-small":
             return self.dim or 1536
+        elif self.dim:
+            # A name not listed above, e.g. a model served behind an
+            # OpenAI-compatible `base_url`: `dim` is its embedding size
+            return self.dim
         else:
-            raise ValueError(f"Unknown model name {self.name}")
+            raise ValueError(
+                f"Unknown model name {self.name}: set `dim` to the model's "
+                "embedding dimension"
+            )
 
     def generate_embeddings(
         self, texts: Union[List[str], "np.ndarray"]
diff --git a/python/python/tests/test_embeddings.py b/python/python/tests/test_embeddings.py
index 32503c04f..c18ddd644 100644
--- a/python/python/tests/test_embeddings.py
+++ b/python/python/tests/test_embeddings.py
@@ -602,6 +602,51 @@ def test_openai_propagates_api_key(monkeypatch):
     assert len(actual.text) > 0
 
 
+@pytest.mark.parametrize("base_url", [None, "http://localhost:11434/v1"])
+def test_openai_compatible_model_uses_dim_for_ndims(base_url):
+    # Unlisted model names use the explicit dimension independently of base_url.
+    openai = get_registry().get("openai")
+    func = openai.create(
+        name="nomic-embed-text",
+        base_url=base_url,
+        dim=768,
+    )
+    assert func.ndims() == 768
+
+    class Schema(LanceModel):
+        text: str = func.SourceField()
+        vector: Vector(func.ndims()) = func.VectorField()
+
+    schema = Schema.to_arrow_schema()
+    assert schema.field("vector").type == pa.list_(pa.float32(), 768)
+
+
+def test_openai_unknown_model_without_dim_raises_clear_error():
+    openai = get_registry().get("openai")
+    func = openai.create(
+        name="nomic-embed-text",
+        base_url="http://localhost:11434/v1",
+    )
+    with pytest.raises(ValueError, match="dim"):
+        func.ndims()
+
+
+@pytest.mark.parametrize(
+    ("name", "dim", "expected_dim"),
+    [
+        ("text-embedding-ada-002", None, 1536),
+        ("text-embedding-ada-002", 512, 1536),
+        ("text-embedding-3-large", None, 3072),
+        ("text-embedding-3-large", 1024, 1024),
+        ("text-embedding-3-small", None, 1536),
+        ("text-embedding-3-small", 512, 512),
+    ],
+)
+def test_openai_known_model_ndims(name, dim, expected_dim):
+    func = get_registry().get("openai").create(name=name, dim=dim)
+    assert func.ndims() == expected_dim
+
+
 @patch("time.sleep")
 def test_openai_no_retry_on_401(mock_sleep):
     """