diff --git a/tests/e2e/nightly/models/test_qwen2_5_vl_32b.py b/tests/e2e/nightly/models/test_qwen2_5_vl_32b.py
index 760f8dee..fe6bbedf 100644
--- a/tests/e2e/nightly/models/test_qwen2_5_vl_32b.py
+++ b/tests/e2e/nightly/models/test_qwen2_5_vl_32b.py
@@ -45,7 +45,7 @@ aisbench_cases = [{
     "dataset_conf": "textvqa/textvqa_gen_base64",
     "max_out_len": 2048,
     "batch_size": 128,
-    "baseline": 76,
+    "baseline": 76.22,
     "temperature": 0,
     "top_k": -1,
     "top_p": 1,
diff --git a/tests/e2e/nightly/models/test_qwen2_5_vl_7b.py b/tests/e2e/nightly/models/test_qwen2_5_vl_7b.py
index bc35ff88..d3a726bf 100644
--- a/tests/e2e/nightly/models/test_qwen2_5_vl_7b.py
+++ b/tests/e2e/nightly/models/test_qwen2_5_vl_7b.py
@@ -45,7 +45,7 @@ aisbench_cases = [{
     "dataset_conf": "textvqa/textvqa_gen_base64",
     "max_out_len": 2048,
     "batch_size": 128,
-    "baseline": 81,
+    "baseline": 82.05,
     "threshold": 5
 }, {
     "case_type": "performance",
diff --git a/tests/e2e/nightly/models/test_qwen3_32b_int8.py b/tests/e2e/nightly/models/test_qwen3_32b_int8.py
index e245f3d7..bbaf863a 100644
--- a/tests/e2e/nightly/models/test_qwen3_32b_int8.py
+++ b/tests/e2e/nightly/models/test_qwen3_32b_int8.py
@@ -58,7 +58,7 @@ aisbench_cases = [{
     "max_out_len": 32768,
     "batch_size": 32,
     "baseline": 83.33,
-    "threshold": 17
+    "threshold": 7
 }, {
     "case_type": "performance",
     "dataset_path": "vllm-ascend/GSM8K-in3500-bs400",