diff --git a/docs/source/prototype.models.rst b/docs/source/prototype.models.rst
new file mode 100644
index 0000000000..489f82e817
--- /dev/null
+++ b/docs/source/prototype.models.rst
@@ -0,0 +1,11 @@
+torchaudio.prototype.models
+===========================
+
+.. py:module:: torchaudio.prototype.models
+.. currentmodule:: torchaudio.prototype.models
+
+conformer_rnnt_base
+~~~~~~~~~~~~~~~~~~~
+
+.. autofunction:: conformer_rnnt_base
+
diff --git a/docs/source/prototype.rst b/docs/source/prototype.rst
index a38550f5af..677e37c4a8 100644
--- a/docs/source/prototype.rst
+++ b/docs/source/prototype.rst
@@ -19,4 +19,5 @@ imported explicitly, e.g.
 .. toctree::
     prototype.io
     prototype.ctc_decoder
+    prototype.models
     prototype.pipelines
diff --git a/torchaudio/prototype/models/rnnt.py b/torchaudio/prototype/models/rnnt.py
index b53335f3af..ec28e82de1 100644
--- a/torchaudio/prototype/models/rnnt.py
+++ b/torchaudio/prototype/models/rnnt.py
@@ -41,6 +41,12 @@ def infer(
 
 
 def conformer_rnnt_base():
+    r"""Builds basic version of Conformer RNN-T model.
+
+        Returns:
+            RNNT:
+                Conformer RNN-T model.
+    """
     encoder = _ConformerTranscriber()
     decoder = _Predictor(
         num_symbols=1024,