diff --git a/qa/L0_backend_python/parameters/response_parameters_test.py b/qa/L0_backend_python/parameters/response_parameters_test.py
index e07bb5eb9f..d8dbbff6cc 100644
--- a/qa/L0_backend_python/parameters/response_parameters_test.py
+++ b/qa/L0_backend_python/parameters/response_parameters_test.py
@@ -166,6 +166,31 @@ def test_setting_response_parameters_decoupled(self):
             output = str(result.as_numpy("OUTPUT")[0][0], encoding="utf-8")
             self.assertEqual(json.dumps(params[i]), output)
 
+    def test_setting_response_parameters_bls(self):
+        model_name = "response_parameters_bls"
+        params = {"bool": False, "int": 2048, "str": "Hello World!"}
+        params_decoupled = [{}, {"bool": True, "int": 10000}, {"str": "?"}]
+        params_str = json.dumps(params)
+        params_decoupled_str = json.dumps(params_decoupled)
+
+        inputs = [
+            grpcclient.InferInput("RESPONSE_PARAMETERS", self._shape, "BYTES"),
+            grpcclient.InferInput(
+                "RESPONSE_PARAMETERS_DECOUPLED", self._shape, "BYTES"
+            ),
+        ]
+        inputs[0].set_data_from_numpy(np.array([[params_str]], dtype=np.object_))
+        inputs[1].set_data_from_numpy(
+            np.array([[params_decoupled_str]], dtype=np.object_)
+        )
+
+        with self._shm_leak_detector.Probe() as shm_probe:
+            with grpcclient.InferenceServerClient(self._server_address_grpc) as client:
+                result = client.infer(model_name, inputs)
+
+        output = str(result.as_numpy("OUTPUT")[0][0], encoding="utf-8")
+        self.assertEqual(output, "True")
+
 
 if __name__ == "__main__":
     unittest.main()
diff --git a/qa/L0_backend_python/parameters/test.sh b/qa/L0_backend_python/parameters/test.sh
index 9d8c86c530..6cf12ccd47 100755
--- a/qa/L0_backend_python/parameters/test.sh
+++ b/qa/L0_backend_python/parameters/test.sh
@@ -39,6 +39,9 @@ mkdir -p models/response_parameters/1 && \
 mkdir -p models/response_parameters_decoupled/1 && \
     cp ../../python_models/response_parameters_decoupled/model.py models/response_parameters_decoupled/1 && \
     cp ../../python_models/response_parameters_decoupled/config.pbtxt models/response_parameters_decoupled
+mkdir -p models/response_parameters_bls/1 && \
+    cp ../../python_models/response_parameters_bls/model.py models/response_parameters_bls/1 && \
+    cp ../../python_models/response_parameters_bls/config.pbtxt models/response_parameters_bls
 
 TEST_LOG="response_parameters_test.log"
 SERVER_LOG="response_parameters_test.server.log"
diff --git a/qa/python_models/response_parameters_bls/config.pbtxt b/qa/python_models/response_parameters_bls/config.pbtxt
new file mode 100644
index 0000000000..1eeb5cd4c2
--- /dev/null
+++ b/qa/python_models/response_parameters_bls/config.pbtxt
@@ -0,0 +1,57 @@
+# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+name: "response_parameters_bls"
+backend: "python"
+max_batch_size: 8
+
+input [
+  {
+    name: "RESPONSE_PARAMETERS"
+    data_type: TYPE_STRING
+    dims: [ 1 ]
+  },
+  {
+    name: "RESPONSE_PARAMETERS_DECOUPLED"
+    data_type: TYPE_STRING
+    dims: [ 1 ]
+  }
+]
+
+output [
+  {
+    name: "OUTPUT"
+    data_type: TYPE_STRING
+    dims: [ 1 ]
+  }
+]
+
+instance_group [
+  {
+    count: 1
+    kind: KIND_CPU
+  }
+]
diff --git a/qa/python_models/response_parameters_bls/model.py b/qa/python_models/response_parameters_bls/model.py
new file mode 100644
index 0000000000..390c14803e
--- /dev/null
+++ b/qa/python_models/response_parameters_bls/model.py
@@ -0,0 +1,106 @@
+# Copyright 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions
+# are met:
+#  * Redistributions of source code must retain the above copyright
+#    notice, this list of conditions and the following disclaimer.
+#  * Redistributions in binary form must reproduce the above copyright
+#    notice, this list of conditions and the following disclaimer in the
+#    documentation and/or other materials provided with the distribution.
+#  * Neither the name of NVIDIA CORPORATION nor the names of its
+#    contributors may be used to endorse or promote products derived
+#    from this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
+# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
+# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+import json
+
+import numpy as np
+import triton_python_backend_utils as pb_utils
+
+
+class TritonPythonModel:
+    """
+    This model (A) is designed to test sending back response parameters when using BLS.
+    It takes one input tensor, which is the RESPONSE_PARAMETERS and uses BLS to
+    call response_parameters model (B). Model B would set RESPONSE_PARAMETERS (with a bit
+    of data massage) as its response parameters. In the end, model A would also set its
+    response parameters from model B's response parameters.
+
+    With above model set up, we can easily test whether the real response parameters are
+    the same as the input response parameters.
+    """
+
+    def execute(self, requests):
+        responses = []
+
+        for request in requests:
+            passed = True
+
+            # test bls response parameters from a regular model
+            res_params_tensor = pb_utils.get_input_tensor_by_name(
+                request, "RESPONSE_PARAMETERS"
+            ).as_numpy()
+            res_params_str = str(res_params_tensor[0][0], encoding="utf-8")
+            res_params = json.loads(res_params_str)
+            bls_input_tensor = pb_utils.Tensor("RESPONSE_PARAMETERS", res_params_tensor)
+            bls_req = pb_utils.InferenceRequest(
+                model_name="response_parameters",
+                inputs=[bls_input_tensor],
+                requested_output_names=["OUTPUT"],
+            )
+            bls_res = bls_req.exec()  # decoupled=False
+            bls_res_params_str = bls_res.parameters()
+            bls_res_params = (
+                json.loads(bls_res_params_str) if bls_res_params_str != "" else {}
+            )
+            passed = passed and bls_res_params == res_params
+
+            # test bls response parameters from a decoupled model
+            res_params_decoupled_tensor = pb_utils.get_input_tensor_by_name(
+                request, "RESPONSE_PARAMETERS_DECOUPLED"
+            ).as_numpy()
+            res_params_decoupled_str = str(
+                res_params_decoupled_tensor[0][0], encoding="utf-8"
+            )
+            res_params_decoupled = json.loads(res_params_decoupled_str)
+            bls_decoupled_input_tensor = pb_utils.Tensor(
+                "RESPONSE_PARAMETERS", res_params_decoupled_tensor
+            )  # response_parameters_decoupled model input name is RESPONSE_PARAMETERS
+            bls_decoupled_req = pb_utils.InferenceRequest(
+                model_name="response_parameters_decoupled",
+                inputs=[bls_decoupled_input_tensor],
+                requested_output_names=["OUTPUT"],
+            )
+            bls_decoupled_res = bls_decoupled_req.exec(decoupled=True)
+            for bls_decoupled_r in bls_decoupled_res:
+                if len(bls_decoupled_r.output_tensors()) == 0:
+                    break  # meaning reached final response
+                bls_decoupled_r_params_str = bls_decoupled_r.parameters()
+                bls_decoupled_r_params = (
+                    json.loads(bls_decoupled_r_params_str)
+                    if bls_decoupled_r_params_str != ""
+                    else {}
+                )
+                passed = passed and bls_decoupled_r_params in res_params_decoupled
+                res_params_decoupled.remove(bls_decoupled_r_params)
+            passed = passed and len(res_params_decoupled) == 0
+
+            output_tensor = pb_utils.Tensor(
+                "OUTPUT", np.array([[str(passed)]], dtype=np.object_)
+            )
+            response = pb_utils.InferenceResponse(output_tensors=[output_tensor])
+            responses.append(response)
+
+        return responses