diff --git a/doc/source/rllib-algorithms.rst b/doc/source/rllib-algorithms.rst
index 8ce7e66ff..575900776 100644
--- a/doc/source/rllib-algorithms.rst
+++ b/doc/source/rllib-algorithms.rst
@@ -274,9 +274,17 @@ Soft Actor Critic (SAC)
 
     SAC architecture (same as DQN)
 
-RLlib's soft-actor critic implementation is ported from the `official SAC repo <https://github.com/rail-berkeley/softlearning>`__ to better integrate with RLlib APIs. Note that SAC has two fields to configure for custom models: ``policy_model`` and ``Q_model``, and currently has no support for non-continuous action distributions. It is also currently *experimental*.
+RLlib's soft-actor critic implementation is ported from the `official SAC repo <https://github.com/rail-berkeley/softlearning>`__ to better integrate with RLlib APIs. Note that SAC has two fields to configure for custom models: ``policy_model`` and ``Q_model``, and currently has no support for non-continuous action distributions.
 
-Tuned examples: `Pendulum-v0 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/regression_tests/pendulum-sac.yaml>`__
+Tuned examples: `Pendulum-v0 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/regression_tests/pendulum-sac.yaml>`__, `HalfCheetah-v3 <https://github.com/ray-project/ray/blob/master/rllib/tuned_examples/halfcheetah-sac.yaml>`__
+
+**MuJoCo results @500k steps:** `more details <https://github.com/ray-project/rl-experiments>`__
+
+=============  ==========  ===================
+MuJoCo env     RLlib SAC   Haarnoja et al SAC
+=============  ==========  ===================
+HalfCheetah    8752        ~9000
+=============  ==========  ===================
 
 **SAC-specific configs** (see also `common configs <rllib-training.html#common-parameters>`__):
 
diff --git a/rllib/agents/sac/sac_model.py b/rllib/agents/sac/sac_model.py
index d72a6063f..93787b535 100644
--- a/rllib/agents/sac/sac_model.py
+++ b/rllib/agents/sac/sac_model.py
@@ -81,7 +81,6 @@ class SACModel(TFModelV2):
             shape=(num_outputs, ), name="model_out")
         self.actions = tf.keras.layers.Input(
             shape=(self.action_dim, ), name="actions")
-
         shift_and_log_scale_diag = tf.keras.Sequential([
             tf.keras.layers.Dense(
                 units=hidden,
@@ -90,10 +89,7 @@ class SACModel(TFModelV2):
             for i, hidden in enumerate(actor_hiddens)
         ] + [
             tf.keras.layers.Dense(
-                units=tfp.layers.MultivariateNormalTriL.params_size(
-                    self.action_dim),
-                activation=None,
-                name="action_out")
+                units=2 * self.action_dim, activation=None, name="action_out")
         ])(self.model_out)
 
         shift, log_scale_diag = tf.keras.layers.Lambda(
diff --git a/rllib/tuned_examples/halfcheetah-sac.yaml b/rllib/tuned_examples/halfcheetah-sac.yaml
new file mode 100644
index 000000000..4669d51bc
--- /dev/null
+++ b/rllib/tuned_examples/halfcheetah-sac.yaml
@@ -0,0 +1,37 @@
+# Our implementation of SAC can reach 9k reward in 400k timesteps
+halfcheetah_sac:
+    env: HalfCheetah-v3
+    run: SAC
+    stop:
+        episode_reward_mean: 9000
+    config:
+        horizon: 1000
+        soft_horizon: False
+        Q_model:
+          hidden_activation: relu
+          hidden_layer_sizes: [256, 256]
+        policy_model:
+          hidden_activation: relu
+          hidden_layer_sizes: [256, 256]
+        tau: 0.005
+        target_entropy: auto
+        no_done_at_end: True
+        n_step: 1
+        sample_batch_size: 1
+        prioritized_replay: False
+        train_batch_size: 256
+        target_network_update_freq: 1
+        timesteps_per_iteration: 1000
+        learning_starts: 10000
+        exploration_enabled: True
+        optimization:
+          actor_learning_rate: 0.0003
+          critic_learning_rate: 0.0003
+          entropy_learning_rate: 0.0003
+        num_workers: 0
+        num_gpus: 0
+        clip_actions: False
+        normalize_actions: True
+        evaluation_interval: 1
+        metrics_smoothing_episodes: 5
+