diff --git a/tests/test_generation.py b/tests/test_generation.py
index 77a73ca15d..c1e28ade3e 100644
--- a/tests/test_generation.py
+++ b/tests/test_generation.py
@@ -29,8 +29,8 @@ def test_eval_gsm8k_api(tmp_path):
     cmd = (
         f"ns eval "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --benchmarks=gsm8k "
         f"    --output_dir={tmp_path} "
         f"    ++max_samples=2 "
@@ -60,12 +60,12 @@ def test_eval_judge_api(tmp_path):
     cmd = (
         f"ns eval "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --benchmarks=math-500 "
         f"    --output_dir={tmp_path} "
-        f"    --judge_model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --judge_server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --judge_model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --judge_server_address=https://integrate.api.nvidia.com/v1 "
         f"    --judge_server_type=openai "
         f"    --judge_generation_type=math_judge "
         f"    --extra_judge_args='++max_concurrent_requests=1 ++inference.timeout=120 ++server.max_retries=1' "
@@ -96,8 +96,8 @@ def test_fail_on_api_key_env_var(tmp_path):
     cmd = (
         f"ns eval "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --benchmarks=gsm8k "
         f"    --output_dir={tmp_path} "
         f"    ++max_samples=2 "
@@ -118,8 +118,8 @@ def test_succeed_on_api_key_env_var(tmp_path):
         f"unset NVIDIA_API_KEY && "
         f"ns eval "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --benchmarks=gsm8k "
         f"    --output_dir={tmp_path} "
         f"    ++max_samples=2 "
@@ -151,8 +151,8 @@ def test_generate_openai_format(tmp_path, format):
     cmd = (
         f"ns generate "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --input_file=/nemo_run/code/tests/data/openai-input-{format}.test "
         f"    --output_dir={tmp_path} "
         f"    ++prompt_format=openai "
@@ -217,12 +217,12 @@ def test_judge_generations_with_structured_output(tmp_path):
     cmd = (
         f"ns eval "
         f"    --server_type=openai "
-        f"    --model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --server_address=https://integrate.api.nvidia.com/v1 "
         f"    --benchmarks=hle "
         f"    --output_dir={tmp_path} "
-        f"    --judge_model=nvidia/nvidia/nemotron-nano-30b-v3 "
-        f"    --judge_server_address=https://inference-api.nvidia.com/v1/ "
+        f"    --judge_model=nvidia/nemotron-3-nano-30b-a3b "
+        f"    --judge_server_address=https://integrate.api.nvidia.com/v1 "
         f"    --judge_server_type=openai "
         f"    --metric_type=hle-aa "
         f'    --extra_judge_args="++structured_output=HLE_JUDGE_AA ++max_concurrent_requests=1 ++inference.timeout=120 ++server.max_retries=1" '