diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 53b7dbf11..d54c2be5b 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -31,7 +31,7 @@ repos:
         args:
           - --select=F401,F821
           - --fix
-        files: ^(benchmark/|docs/|examples/|python/sglang/|sgl-router/py_*)
+        files: ^(benchmark/|docs/|examples/|python/sglang/|sgl-router/py_*|test/)
         exclude: |
           (?x)^(
           .*/__init__\.py$|
diff --git a/test/lang/test_anthropic_backend.py b/test/lang/test_anthropic_backend.py
index dc8f0b17e..be4d5921c 100644
--- a/test/lang/test_anthropic_backend.py
+++ b/test/lang/test_anthropic_backend.py
@@ -1,4 +1,3 @@
-import json
 import unittest
 
 from sglang import Anthropic, set_default_backend
diff --git a/test/lang/test_litellm_backend.py b/test/lang/test_litellm_backend.py
index 74c3a187a..8fcc69cf7 100644
--- a/test/lang/test_litellm_backend.py
+++ b/test/lang/test_litellm_backend.py
@@ -1,4 +1,3 @@
-import json
 import unittest
 
 from sglang import LiteLLM, set_default_backend
diff --git a/test/lang/test_separate_reasoning.py b/test/lang/test_separate_reasoning.py
index 1709cecfd..20a8237eb 100644
--- a/test/lang/test_separate_reasoning.py
+++ b/test/lang/test_separate_reasoning.py
@@ -7,7 +7,7 @@ python3 -m unittest test/lang/test_separate_reasoning.py
 
 import unittest
 
-from sglang import assistant, gen, separate_reasoning, user
+from sglang import gen, separate_reasoning
 from sglang.lang.ir import SglExprList, SglSeparateReasoning
 from sglang.test.test_utils import CustomTestCase
 
diff --git a/test/lang/test_separate_reasoning_execution.py b/test/lang/test_separate_reasoning_execution.py
index 481488f6a..9dd4b4134 100644
--- a/test/lang/test_separate_reasoning_execution.py
+++ b/test/lang/test_separate_reasoning_execution.py
@@ -6,11 +6,9 @@ python3 -m unittest test/lang/test_separate_reasoning_execution.py
 """
 
 import threading
-import time
 import unittest
 from unittest.mock import MagicMock, patch
 
-from sglang import assistant, gen, separate_reasoning, user
 from sglang.lang.interpreter import StreamExecutor
 from sglang.lang.ir import SglGen, SglSeparateReasoning
 from sglang.test.test_utils import CustomTestCase
diff --git a/test/srt/ascend/test_ascend_deepep.py b/test/srt/ascend/test_ascend_deepep.py
index eb05f559e..d6a94890a 100644
--- a/test/srt/ascend/test_ascend_deepep.py
+++ b/test/srt/ascend/test_ascend_deepep.py
@@ -6,7 +6,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,
diff --git a/test/srt/ascend/test_ascend_deepseek_mtp.py b/test/srt/ascend/test_ascend_deepseek_mtp.py
index d235ee59b..e149c4735 100644
--- a/test/srt/ascend/test_ascend_deepseek_mtp.py
+++ b/test/srt/ascend/test_ascend_deepseek_mtp.py
@@ -6,7 +6,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,
diff --git a/test/srt/ascend/test_ascend_tp4_bf16.py b/test/srt/ascend/test_ascend_tp4_bf16.py
index bb7d90e4f..e1feac417 100644
--- a/test/srt/ascend/test_ascend_tp4_bf16.py
+++ b/test/srt/ascend/test_ascend_tp4_bf16.py
@@ -5,7 +5,6 @@ from urllib.parse import urlparse
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
     is_in_ci,
diff --git a/test/srt/cpu/test_activation.py b/test/srt/cpu/test_activation.py
index 05447bc80..ffb068fc3 100644
--- a/test/srt/cpu/test_activation.py
+++ b/test/srt/cpu/test_activation.py
@@ -1,9 +1,7 @@
 import itertools
 import unittest
 
-import sgl_kernel
 import torch
-import torch.nn.functional as F
 from utils import GeluAndMul, SiluAndMul, precision
 
 from sglang.srt.server_args import ServerArgs, set_global_server_args_for_scheduler
diff --git a/test/srt/cpu/test_binding.py b/test/srt/cpu/test_binding.py
index d3cc329af..94dd5b11b 100644
--- a/test/srt/cpu/test_binding.py
+++ b/test/srt/cpu/test_binding.py
@@ -1,7 +1,6 @@
 import re
 import unittest
 
-import sgl_kernel
 import torch
 
 kernel = torch.ops.sgl_kernel
diff --git a/test/srt/cpu/test_comm.py b/test/srt/cpu/test_comm.py
index 60e7194bd..0bd187a03 100644
--- a/test/srt/cpu/test_comm.py
+++ b/test/srt/cpu/test_comm.py
@@ -5,11 +5,9 @@ import traceback
 import unittest
 from multiprocessing import Process
 
-import sgl_kernel
 import torch
 import torch.distributed as dist
 import torch.multiprocessing as mp
-from utils import precision
 
 from sglang.test.test_utils import CustomTestCase, find_available_port
 
diff --git a/test/srt/cpu/test_decode.py b/test/srt/cpu/test_decode.py
index a7c5dd755..aaf5f5d47 100644
--- a/test/srt/cpu/test_decode.py
+++ b/test/srt/cpu/test_decode.py
@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 
diff --git a/test/srt/cpu/test_extend.py b/test/srt/cpu/test_extend.py
index 9c6f5b394..7277050c2 100644
--- a/test/srt/cpu/test_extend.py
+++ b/test/srt/cpu/test_extend.py
@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 
diff --git a/test/srt/cpu/test_gemm.py b/test/srt/cpu/test_gemm.py
index a9fe5066a..b9ec1e7bf 100644
--- a/test/srt/cpu/test_gemm.py
+++ b/test/srt/cpu/test_gemm.py
@@ -2,7 +2,6 @@ import itertools
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
 import torch.nn as nn
 from utils import (
diff --git a/test/srt/cpu/test_mla.py b/test/srt/cpu/test_mla.py
index 1f0718d7a..3caa109d5 100644
--- a/test/srt/cpu/test_mla.py
+++ b/test/srt/cpu/test_mla.py
@@ -1,7 +1,5 @@
-import itertools
 import unittest
 
-import sgl_kernel
 import torch
 from torch.nn.functional import scaled_dot_product_attention
 from utils import precision
diff --git a/test/srt/cpu/test_moe.py b/test/srt/cpu/test_moe.py
index 96eb28020..7babd5167 100644
--- a/test/srt/cpu/test_moe.py
+++ b/test/srt/cpu/test_moe.py
@@ -3,7 +3,6 @@ import math
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
 
 kernel = torch.ops.sgl_kernel
diff --git a/test/srt/cpu/test_norm.py b/test/srt/cpu/test_norm.py
index 75bacb198..ebae9eff6 100644
--- a/test/srt/cpu/test_norm.py
+++ b/test/srt/cpu/test_norm.py
@@ -2,7 +2,6 @@ import itertools
 import unittest
 from typing import Optional, Tuple, Union
 
-import sgl_kernel
 import torch
 from utils import make_non_contiguous, precision
 
diff --git a/test/srt/cpu/test_qkv_proj_with_rope.py b/test/srt/cpu/test_qkv_proj_with_rope.py
index dc90cc559..b0b22d3bf 100644
--- a/test/srt/cpu/test_qkv_proj_with_rope.py
+++ b/test/srt/cpu/test_qkv_proj_with_rope.py
@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from utils import (
     convert_weight,
diff --git a/test/srt/cpu/test_rope.py b/test/srt/cpu/test_rope.py
index 35c622e54..8c1dfe9aa 100644
--- a/test/srt/cpu/test_rope.py
+++ b/test/srt/cpu/test_rope.py
@@ -1,6 +1,5 @@
 import unittest
 
-import sgl_kernel
 import torch
 from utils import precision
 
diff --git a/test/srt/cpu/test_shared_expert.py b/test/srt/cpu/test_shared_expert.py
index 6086fed65..358709a6a 100644
--- a/test/srt/cpu/test_shared_expert.py
+++ b/test/srt/cpu/test_shared_expert.py
@@ -3,9 +3,7 @@ import math
 import unittest
 
 # TODO: use interface in cpu.py
-import sgl_kernel
 import torch
-import torch.nn as nn
 from utils import (
     BLOCK_K,
     BLOCK_N,
diff --git a/test/srt/cpu/test_topk.py b/test/srt/cpu/test_topk.py
index 4b4ce21ae..9f3dfc1b4 100644
--- a/test/srt/cpu/test_topk.py
+++ b/test/srt/cpu/test_topk.py
@@ -1,9 +1,6 @@
-import itertools
 import unittest
 
-import sgl_kernel
 import torch
-from utils import precision
 
 from sglang.srt.layers.moe.topk import (
     biased_grouped_topk_impl as native_biased_grouped_topk,
diff --git a/test/srt/entrypoints/http_server/test_abort_request.py b/test/srt/entrypoints/http_server/test_abort_request.py
index 7a366f244..7d9a60a1b 100644
--- a/test/srt/entrypoints/http_server/test_abort_request.py
+++ b/test/srt/entrypoints/http_server/test_abort_request.py
@@ -8,7 +8,6 @@ Run with:
 import threading
 import time
 import unittest
-from typing import Optional
 
 import requests
 
diff --git a/test/srt/ep/test_deepep_intranode.py b/test/srt/ep/test_deepep_intranode.py
index 97acd000c..ef365b010 100644
--- a/test/srt/ep/test_deepep_intranode.py
+++ b/test/srt/ep/test_deepep_intranode.py
@@ -1,6 +1,5 @@
 # Copy from deepseek-ai/DeepEP/tests/test_intranode.py
 
-import os
 import time
 
 # noinspection PyUnresolvedReferences
diff --git a/test/srt/ep/test_deepep_small.py b/test/srt/ep/test_deepep_small.py
index 05aefe79a..4417affb9 100644
--- a/test/srt/ep/test_deepep_small.py
+++ b/test/srt/ep/test_deepep_small.py
@@ -317,7 +317,6 @@ class TestMTP(CustomTestCase):
 class TestMTPWithTBO(CustomTestCase):
     @classmethod
     def setUpClass(cls):
-        import os
 
         cls.model = DEFAULT_MODEL_NAME_FOR_TEST_MLA
         cls.base_url = DEFAULT_URL_FOR_TEST
diff --git a/test/srt/ep/test_eplb.py b/test/srt/ep/test_eplb.py
index 748dd39c8..0da365f33 100755
--- a/test/srt/ep/test_eplb.py
+++ b/test/srt/ep/test_eplb.py
@@ -1,4 +1,3 @@
-import os
 import tempfile
 import unittest
 from pathlib import Path
diff --git a/test/srt/ep/test_moe_deepep.py b/test/srt/ep/test_moe_deepep.py
index 00f7cd59b..495a3aeff 100644
--- a/test/srt/ep/test_moe_deepep.py
+++ b/test/srt/ep/test_moe_deepep.py
@@ -1,5 +1,4 @@
 import json
-import os
 import unittest
 from types import SimpleNamespace
 
diff --git a/test/srt/experiment_runner.py b/test/srt/experiment_runner.py
index f32f61d3b..f6af05623 100644
--- a/test/srt/experiment_runner.py
+++ b/test/srt/experiment_runner.py
@@ -8,8 +8,7 @@ import threading
 import time
 from dataclasses import dataclass
 from datetime import datetime
-from pathlib import Path
-from typing import Dict, List, Optional, Tuple
+from typing import List, Optional, Tuple
 
 import psutil
 import requests
diff --git a/test/srt/function_call/test_json_schema_constraint.py b/test/srt/function_call/test_json_schema_constraint.py
index 962d97ea2..b1e6195e7 100644
--- a/test/srt/function_call/test_json_schema_constraint.py
+++ b/test/srt/function_call/test_json_schema_constraint.py
@@ -12,7 +12,6 @@ from sglang.srt.entrypoints.openai.protocol import (
     ToolChoice,
     ToolChoiceFuncName,
 )
-from sglang.srt.function_call.function_call_parser import FunctionCallParser
 from sglang.srt.function_call.utils import (
     _get_tool_schema_defs,
     get_json_schema_constraint,
diff --git a/test/srt/hicache/test_hicache_storage_3fs_backend.py b/test/srt/hicache/test_hicache_storage_3fs_backend.py
index d3dea5117..c91772881 100644
--- a/test/srt/hicache/test_hicache_storage_3fs_backend.py
+++ b/test/srt/hicache/test_hicache_storage_3fs_backend.py
@@ -6,13 +6,10 @@ Usage:
 
 import json
 import os
-import time
 import unittest
-from types import SimpleNamespace
 
 from test_hicache_storage_file_backend import HiCacheStorageBaseMixin
 
-from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import CustomTestCase
 
 
diff --git a/test/srt/layers/moe/test_moe_runners.py b/test/srt/layers/moe/test_moe_runners.py
index f325a0212..4b4047d4b 100644
--- a/test/srt/layers/moe/test_moe_runners.py
+++ b/test/srt/layers/moe/test_moe_runners.py
@@ -4,11 +4,9 @@ from types import SimpleNamespace
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
-    DEFAULT_AWQ_MOE_MODEL_NAME_FOR_TEST,
     DEFAULT_MODEL_NAME_FOR_TEST_FP8_WITH_MOE,
     DEFAULT_MODEL_NAME_FOR_TEST_MOE_NVFP4,
     DEFAULT_MODEL_NAME_FOR_TEST_MXFP4_WITH_MOE,
-    DEFAULT_MODEL_NAME_FOR_TEST_W8A8_WITH_MOE,
     DEFAULT_SMALL_MOE_MODEL_NAME_FOR_TEST_CHAT,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
diff --git a/test/srt/lora/test_lora.py b/test/srt/lora/test_lora.py
index 3ab7b624d..aff0b3a79 100644
--- a/test/srt/lora/test_lora.py
+++ b/test/srt/lora/test_lora.py
@@ -18,7 +18,6 @@ import random
 import unittest
 from typing import List
 
-import torch
 from utils import (
     ALL_OTHER_MULTI_LORA_MODELS,
     CI_MULTI_LORA_MODELS,
diff --git a/test/srt/lora/test_lora_openai_api.py b/test/srt/lora/test_lora_openai_api.py
index 4f5ac5303..649e4424e 100644
--- a/test/srt/lora/test_lora_openai_api.py
+++ b/test/srt/lora/test_lora_openai_api.py
@@ -6,7 +6,7 @@ that enables OpenAI-compatible LoRA adapter selection.
 """
 
 import unittest
-from unittest.mock import MagicMock, Mock
+from unittest.mock import MagicMock
 
 from sglang.srt.entrypoints.openai.serving_base import OpenAIServingBase
 from sglang.srt.server_args import ServerArgs
diff --git a/test/srt/lora/test_lora_radix_cache.py b/test/srt/lora/test_lora_radix_cache.py
index 2faacf930..8dcd09b79 100644
--- a/test/srt/lora/test_lora_radix_cache.py
+++ b/test/srt/lora/test_lora_radix_cache.py
@@ -13,13 +13,11 @@
 # ==============================================================================
 
 import multiprocessing as mp
-import random
 import unittest
 
 import torch
-from utils import CI_MULTI_LORA_MODELS, DEFAULT_PROMPTS, run_lora_test_one_by_one
+from utils import CI_MULTI_LORA_MODELS, run_lora_test_one_by_one
 
-from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase
 
 PROMPTS = [
diff --git a/test/srt/lora/utils.py b/test/srt/lora/utils.py
index 95089d33c..27e6b72c7 100644
--- a/test/srt/lora/utils.py
+++ b/test/srt/lora/utils.py
@@ -14,7 +14,7 @@
 
 import dataclasses
 import random
-from typing import List, Optional
+from typing import List
 
 import torch
 
diff --git a/test/srt/models/test_clip_models.py b/test/srt/models/test_clip_models.py
index 8a79656d0..e537a7466 100644
--- a/test/srt/models/test_clip_models.py
+++ b/test/srt/models/test_clip_models.py
@@ -16,10 +16,8 @@ import multiprocessing as mp
 import unittest
 
 import torch
-from transformers import AutoProcessor
 
-from sglang.srt.utils import load_image
-from sglang.test.runners import DEFAULT_PROMPTS, HFRunner, SRTRunner
+from sglang.test.runners import HFRunner, SRTRunner
 from sglang.test.test_utils import get_similarities
 
 TEXTS = "two Subway Series sandwiches with meats, cheese, lettuce, tomatoes, and onions on a black background, accompanied by the Subway Series logo, highlighting a new sandwich series."
diff --git a/test/srt/models/test_cross_encoder_models.py b/test/srt/models/test_cross_encoder_models.py
index 93edc3fa1..6ff963dd4 100644
--- a/test/srt/models/test_cross_encoder_models.py
+++ b/test/srt/models/test_cross_encoder_models.py
@@ -3,7 +3,6 @@ import random
 import unittest
 
 import torch
-from transformers import AutoConfig, AutoTokenizer
 
 from sglang.test.runners import TEST_RERANK_QUERY_DOCS, HFRunner, SRTRunner
 from sglang.test.test_utils import CustomTestCase, is_in_ci
diff --git a/test/srt/models/test_falcon_h1_models.py b/test/srt/models/test_falcon_h1_models.py
index cb32a7ef1..1706cc859 100644
--- a/test/srt/models/test_falcon_h1_models.py
+++ b/test/srt/models/test_falcon_h1_models.py
@@ -1,4 +1,3 @@
-import unittest
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
diff --git a/test/srt/models/test_generation_models.py b/test/srt/models/test_generation_models.py
index 4aab2246e..43770d5a1 100644
--- a/test/srt/models/test_generation_models.py
+++ b/test/srt/models/test_generation_models.py
@@ -22,7 +22,6 @@ To test a specific model locally:
 import dataclasses
 import multiprocessing as mp
 import os
-import random
 import unittest
 from typing import List
 
diff --git a/test/srt/models/test_llama4_models.py b/test/srt/models/test_llama4_models.py
index 3835ca8d7..cb0c57604 100644
--- a/test/srt/models/test_llama4_models.py
+++ b/test/srt/models/test_llama4_models.py
@@ -1,4 +1,3 @@
-import random
 import unittest
 from types import SimpleNamespace
 
diff --git a/test/srt/nightly/test_gsm8k_eval_amd.py b/test/srt/nightly/test_gsm8k_eval_amd.py
index 232fde507..f27998632 100644
--- a/test/srt/nightly/test_gsm8k_eval_amd.py
+++ b/test/srt/nightly/test_gsm8k_eval_amd.py
@@ -2,7 +2,6 @@ import json
 import os
 import unittest
 import warnings
-from datetime import datetime
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
diff --git a/test/srt/openai_server/basic/test_openai_embedding.py b/test/srt/openai_server/basic/test_openai_embedding.py
index d198b1a7f..21da5dd84 100644
--- a/test/srt/openai_server/basic/test_openai_embedding.py
+++ b/test/srt/openai_server/basic/test_openai_embedding.py
@@ -1,8 +1,6 @@
 import json
-import os
 import unittest
 
-import numpy as np
 import openai
 
 from sglang.srt.utils import kill_process_tree
diff --git a/test/srt/openai_server/basic/test_openai_server.py b/test/srt/openai_server/basic/test_openai_server.py
index c32ef9e90..9b09fbfaf 100644
--- a/test/srt/openai_server/basic/test_openai_server.py
+++ b/test/srt/openai_server/basic/test_openai_server.py
@@ -7,12 +7,10 @@ python3 -m unittest openai_server.basic.test_openai_server.TestOpenAIServer.test
 
 import json
 import random
-import re
 import unittest
 from concurrent.futures import ThreadPoolExecutor
 from typing import Optional
 
-import numpy as np
 import openai
 import requests
 
diff --git a/test/srt/openai_server/basic/test_protocol.py b/test/srt/openai_server/basic/test_protocol.py
index fbf1e3971..289b6638d 100644
--- a/test/srt/openai_server/basic/test_protocol.py
+++ b/test/srt/openai_server/basic/test_protocol.py
@@ -13,53 +13,19 @@
 # ==============================================================================
 """Tests for OpenAI API protocol models"""
 
-import json
-import time
 import unittest
-from typing import Dict, List, Optional
+from typing import List, Optional
 
 from pydantic import BaseModel, Field, ValidationError
 
 from sglang.srt.entrypoints.openai.protocol import (
-    BatchRequest,
-    BatchResponse,
-    ChatCompletionMessageContentImagePart,
-    ChatCompletionMessageContentTextPart,
     ChatCompletionRequest,
     ChatCompletionResponse,
     ChatCompletionResponseChoice,
-    ChatCompletionResponseStreamChoice,
-    ChatCompletionStreamResponse,
-    ChatCompletionTokenLogprob,
     ChatMessage,
-    ChoiceLogprobs,
     CompletionRequest,
-    CompletionResponse,
-    CompletionResponseChoice,
-    DeltaMessage,
-    EmbeddingObject,
-    EmbeddingRequest,
-    EmbeddingResponse,
-    ErrorResponse,
-    FileDeleteResponse,
-    FileRequest,
-    FileResponse,
-    Function,
-    FunctionResponse,
-    JsonSchemaResponseFormat,
-    LogProbs,
     ModelCard,
     ModelList,
-    MultimodalEmbeddingInput,
-    ResponseFormat,
-    ScoringRequest,
-    ScoringResponse,
-    StreamOptions,
-    StructuralTagResponseFormat,
-    Tool,
-    ToolCall,
-    ToolChoice,
-    TopLogprob,
     UsageInfo,
 )
 
diff --git a/test/srt/openai_server/basic/test_serving_completions.py b/test/srt/openai_server/basic/test_serving_completions.py
index 022ba9ad1..93aba834e 100644
--- a/test/srt/openai_server/basic/test_serving_completions.py
+++ b/test/srt/openai_server/basic/test_serving_completions.py
@@ -6,7 +6,7 @@ Run with:
 
 import unittest
 from typing import Optional
-from unittest.mock import AsyncMock, Mock, patch
+from unittest.mock import AsyncMock, Mock
 
 from sglang.srt.entrypoints.openai.protocol import CompletionRequest
 from sglang.srt.entrypoints.openai.serving_completions import OpenAIServingCompletion
diff --git a/test/srt/openai_server/basic/test_serving_embedding.py b/test/srt/openai_server/basic/test_serving_embedding.py
index 47f8330b7..2621eb521 100644
--- a/test/srt/openai_server/basic/test_serving_embedding.py
+++ b/test/srt/openai_server/basic/test_serving_embedding.py
@@ -10,7 +10,6 @@ from fastapi import Request
 
 from sglang.srt.entrypoints.openai.protocol import (
     EmbeddingRequest,
-    EmbeddingResponse,
     MultimodalEmbeddingInput,
 )
 from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbedding
diff --git a/test/srt/openai_server/features/test_cache_report.py b/test/srt/openai_server/features/test_cache_report.py
index 939556993..6a5f7bd8a 100644
--- a/test/srt/openai_server/features/test_cache_report.py
+++ b/test/srt/openai_server/features/test_cache_report.py
@@ -1,4 +1,3 @@
-import asyncio
 import unittest
 
 import openai
diff --git a/test/srt/openai_server/features/test_enable_thinking.py b/test/srt/openai_server/features/test_enable_thinking.py
index 5e03d17de..5687c68e2 100644
--- a/test/srt/openai_server/features/test_enable_thinking.py
+++ b/test/srt/openai_server/features/test_enable_thinking.py
@@ -6,18 +6,12 @@ python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinki
 python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_stream_chat_completion_without_reasoning
 """
 
-import asyncio
 import json
-import os
-import sys
-import time
 import unittest
 
-import openai
 import requests
 
 from sglang.srt.utils import kill_process_tree
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_ENABLE_THINKING_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
diff --git a/test/srt/openai_server/features/test_openai_server_hidden_states.py b/test/srt/openai_server/features/test_openai_server_hidden_states.py
index bb066e691..6544a9a87 100644
--- a/test/srt/openai_server/features/test_openai_server_hidden_states.py
+++ b/test/srt/openai_server/features/test_openai_server_hidden_states.py
@@ -1,19 +1,13 @@
-import json
-import re
-import time
 import unittest
 from abc import ABC
 
-import numpy as np
 import openai
-import torch
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
-    DEFAULT_SMALL_EMBEDDING_MODEL_NAME_FOR_TEST,
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
diff --git a/test/srt/openai_server/features/test_reasoning_content.py b/test/srt/openai_server/features/test_reasoning_content.py
index 04e5160a7..d605a24f6 100644
--- a/test/srt/openai_server/features/test_reasoning_content.py
+++ b/test/srt/openai_server/features/test_reasoning_content.py
@@ -9,11 +9,9 @@ python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningC
 python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningContentStartup.test_streaming
 """
 
-import json
 import unittest
 
 import openai
-import requests
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (
diff --git a/test/srt/openai_server/function_call/test_openai_function_calling.py b/test/srt/openai_server/function_call/test_openai_function_calling.py
index fe5a49728..b54f90fc9 100644
--- a/test/srt/openai_server/function_call/test_openai_function_calling.py
+++ b/test/srt/openai_server/function_call/test_openai_function_calling.py
@@ -1,5 +1,4 @@
 import json
-import time
 import unittest
 
 import openai
diff --git a/test/srt/quant/test_awq.py b/test/srt/quant/test_awq.py
index c32b9f979..f6b74669c 100644
--- a/test/srt/quant/test_awq.py
+++ b/test/srt/quant/test_awq.py
@@ -1,8 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
diff --git a/test/srt/quant/test_triton_scaled_mm.py b/test/srt/quant/test_triton_scaled_mm.py
index dafde83be..72a0bbf31 100644
--- a/test/srt/quant/test_triton_scaled_mm.py
+++ b/test/srt/quant/test_triton_scaled_mm.py
@@ -1,4 +1,3 @@
-import itertools
 import unittest
 from typing import Optional
 
diff --git a/test/srt/rl/test_update_weights_from_distributed.py b/test/srt/rl/test_update_weights_from_distributed.py
index 37782c397..f2c6e5131 100644
--- a/test/srt/rl/test_update_weights_from_distributed.py
+++ b/test/srt/rl/test_update_weights_from_distributed.py
@@ -22,7 +22,6 @@ import unittest
 import numpy as np
 import requests
 import torch
-import torch.distributed as dist
 import torch.multiprocessing as mp
 from transformers import AutoModelForCausalLM
 
diff --git a/test/srt/rl/test_verl_engine_2_gpu.py b/test/srt/rl/test_verl_engine_2_gpu.py
index 39b2e6887..fbd66f062 100644
--- a/test/srt/rl/test_verl_engine_2_gpu.py
+++ b/test/srt/rl/test_verl_engine_2_gpu.py
@@ -19,7 +19,6 @@ from torch.distributed.fsdp.api import (
 from transformers import AutoModelForCausalLM
 
 from sglang.srt.entrypoints.verl_engine import VerlEngine
-from sglang.srt.utils import is_port_available
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.runners import (
     HFRunner,
diff --git a/test/srt/rl/test_verl_engine_4_gpu.py b/test/srt/rl/test_verl_engine_4_gpu.py
index fb137cab4..a6fce4f88 100644
--- a/test/srt/rl/test_verl_engine_4_gpu.py
+++ b/test/srt/rl/test_verl_engine_4_gpu.py
@@ -19,7 +19,6 @@ from torch.distributed.fsdp.api import (
 from transformers import AutoModelForCausalLM
 
 from sglang.srt.entrypoints.verl_engine import VerlEngine
-from sglang.srt.utils import is_port_available
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.runners import (
     HFRunner,
@@ -125,7 +124,7 @@ class TestVerlEngine(CustomTestCase):
         if is_in_ci():
             return
 
-        for index, model_info in enumerate(ALL_OTHER_MODELS):
+        for index, model_info in enumerate(ALL_MODELS):
             self.assert_fragment_e2e_execution(index=index, **model_info)
 
     # def test_adhoc(self):
diff --git a/test/srt/test_abort.py b/test/srt/test_abort.py
index f50ab4037..94256e558 100644
--- a/test/srt/test_abort.py
+++ b/test/srt/test_abort.py
@@ -1,6 +1,4 @@
-import json
 import multiprocessing
-import os
 import time
 import unittest
 from concurrent.futures import ThreadPoolExecutor, as_completed
diff --git a/test/srt/test_async_dynamic_batch_tokenizer.py b/test/srt/test_async_dynamic_batch_tokenizer.py
index 930e23e54..f5d50ab56 100644
--- a/test/srt/test_async_dynamic_batch_tokenizer.py
+++ b/test/srt/test_async_dynamic_batch_tokenizer.py
@@ -8,7 +8,7 @@ including batch efficiency, timeout handling, and error cases.
 import asyncio
 import logging
 import time
-from unittest.mock import AsyncMock, Mock, patch
+from unittest.mock import Mock
 
 import pytest
 from transformers import AutoTokenizer
diff --git a/test/srt/test_bench_one_batch.py b/test/srt/test_bench_one_batch.py
index 7ec33a559..8d14cd0b7 100644
--- a/test/srt/test_bench_one_batch.py
+++ b/test/srt/test_bench_one_batch.py
@@ -1,4 +1,3 @@
-import os
 import unittest
 
 from sglang.test.test_utils import (
diff --git a/test/srt/test_bench_serving.py b/test/srt/test_bench_serving.py
index 747794609..d78c16ba8 100644
--- a/test/srt/test_bench_serving.py
+++ b/test/srt/test_bench_serving.py
@@ -4,7 +4,6 @@ import unittest
 
 import requests
 
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
diff --git a/test/srt/test_bnb.py b/test/srt/test_bnb.py
index 4505b20cf..4328d56be 100644
--- a/test/srt/test_bnb.py
+++ b/test/srt/test_bnb.py
@@ -4,19 +4,12 @@ python3 -m unittest test_bnb.TestVisionModel.test_vlm
 python3 -m unittest test_bnb.TestLanguageModel.test_mmlu
 """
 
-import io
-import json
 import multiprocessing as mp
-import os
-import unittest
+import random
 from concurrent.futures import ThreadPoolExecutor
 from types import SimpleNamespace
 
-import numpy as np
 import openai
-import pybase64
-import requests
-from PIL import Image
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
diff --git a/test/srt/test_config_integration.py b/test/srt/test_config_integration.py
index ea13b8d6c..6b3992c89 100644
--- a/test/srt/test_config_integration.py
+++ b/test/srt/test_config_integration.py
@@ -4,7 +4,6 @@ Test script to verify SGLang config file integration.
 
 import os
 import tempfile
-from pathlib import Path
 
 import pytest
 import yaml
diff --git a/test/srt/test_cpp_radix_cache.py b/test/srt/test_cpp_radix_cache.py
index 2e6baa653..b2146beaf 100644
--- a/test/srt/test_cpp_radix_cache.py
+++ b/test/srt/test_cpp_radix_cache.py
@@ -1,4 +1,3 @@
-import os
 import unittest
 from types import SimpleNamespace
 
diff --git a/test/srt/test_create_kvindices.py b/test/srt/test_create_kvindices.py
index 4196eb290..7e63fd823 100644
--- a/test/srt/test_create_kvindices.py
+++ b/test/srt/test_create_kvindices.py
@@ -1,4 +1,3 @@
-import itertools
 import unittest
 
 import numpy as np
diff --git a/test/srt/test_deepseek_chat_templates.py b/test/srt/test_deepseek_chat_templates.py
index 6c56816c8..a2c1c8ecd 100644
--- a/test/srt/test_deepseek_chat_templates.py
+++ b/test/srt/test_deepseek_chat_templates.py
@@ -6,7 +6,6 @@ both dict and string types for tool['function']['arguments'] without double-esca
 addressing issue #11700.
 """
 
-import json
 import os
 import unittest
 
diff --git a/test/srt/test_deepseek_v32_nsabackend.py b/test/srt/test_deepseek_v32_nsabackend.py
index cd7702f28..205983ab3 100644
--- a/test/srt/test_deepseek_v32_nsabackend.py
+++ b/test/srt/test_deepseek_v32_nsabackend.py
@@ -3,7 +3,6 @@ from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
-from sglang.test.send_one import BenchArgs, send_one_prompt
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
diff --git a/test/srt/test_deepseek_v3_basic.py b/test/srt/test_deepseek_v3_basic.py
index 989192e19..e8c4b2d08 100644
--- a/test/srt/test_deepseek_v3_basic.py
+++ b/test/srt/test_deepseek_v3_basic.py
@@ -1,8 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.send_one import BenchArgs, send_one_prompt
diff --git a/test/srt/test_deepseek_v3_cutedsl_4gpu.py b/test/srt/test_deepseek_v3_cutedsl_4gpu.py
index a72a19027..260575ef0 100644
--- a/test/srt/test_deepseek_v3_cutedsl_4gpu.py
+++ b/test/srt/test_deepseek_v3_cutedsl_4gpu.py
@@ -2,8 +2,6 @@ import os
 import unittest
 from types import SimpleNamespace
 
-import requests
-
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
diff --git a/test/srt/test_deterministic.py b/test/srt/test_deterministic.py
index f0fcc426b..4d6c8e608 100644
--- a/test/srt/test_deterministic.py
+++ b/test/srt/test_deterministic.py
@@ -9,19 +9,10 @@ test into unit tests so that's easily reproducible in CI.
 
 import unittest
 
-from sglang.srt.utils import kill_process_tree
-from sglang.test.test_deterministic import BenchArgs, test_deterministic
 from sglang.test.test_deterministic_utils import (
     COMMON_SERVER_ARGS,
-    DEFAULT_MODEL,
     TestDeterministicBase,
 )
-from sglang.test.test_utils import (
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
-    DEFAULT_URL_FOR_TEST,
-    CustomTestCase,
-    popen_launch_server,
-)
 
 
 class TestFlashinferDeterministic(TestDeterministicBase):
diff --git a/test/srt/test_disaggregation_hybrid_attention.py b/test/srt/test_disaggregation_hybrid_attention.py
index 34ed29c72..83ebed93f 100644
--- a/test/srt/test_disaggregation_hybrid_attention.py
+++ b/test/srt/test_disaggregation_hybrid_attention.py
@@ -1,8 +1,6 @@
-import os
 import unittest
 from types import SimpleNamespace
 
-from sglang.srt.environ import envs
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_disaggregation_utils import TestDisaggregationBase
 from sglang.test.test_utils import (
diff --git a/test/srt/test_eval_accuracy_large.py b/test/srt/test_eval_accuracy_large.py
index 99304dfde..efb202463 100644
--- a/test/srt/test_eval_accuracy_large.py
+++ b/test/srt/test_eval_accuracy_large.py
@@ -3,8 +3,6 @@ Usage:
 python -m unittest test_eval_accuracy_large.TestEvalAccuracyLarge.test_mmlu
 """
 
-import os
-import time
 import unittest
 from types import SimpleNamespace
 
diff --git a/test/srt/test_expert_distribution.py b/test/srt/test_expert_distribution.py
index aeb989941..c3ebd0762 100755
--- a/test/srt/test_expert_distribution.py
+++ b/test/srt/test_expert_distribution.py
@@ -1,4 +1,3 @@
-import os
 import tempfile
 import unittest
 from pathlib import Path
diff --git a/test/srt/test_flash_attention_4.py b/test/srt/test_flash_attention_4.py
index 2be593b4a..4322263c4 100644
--- a/test/srt/test_flash_attention_4.py
+++ b/test/srt/test_flash_attention_4.py
@@ -1,7 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-from sglang.srt.environ import envs
 from sglang.srt.utils import get_device_sm, kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
 from sglang.test.test_utils import (
diff --git a/test/srt/test_flashmla.py b/test/srt/test_flashmla.py
index 07d19f9f9..e9c69e5d5 100644
--- a/test/srt/test_flashmla.py
+++ b/test/srt/test_flashmla.py
@@ -11,15 +11,12 @@ import torch
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
-from sglang.test.send_one import BenchArgs, send_one_prompt
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST_MLA,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
-    is_in_ci,
     popen_launch_server,
-    write_github_step_summary,
 )
 
 
diff --git a/test/srt/test_fp4_moe.py b/test/srt/test_fp4_moe.py
index ac42fa860..306a331fe 100644
--- a/test/srt/test_fp4_moe.py
+++ b/test/srt/test_fp4_moe.py
@@ -5,12 +5,11 @@ import pytest
 import torch
 from flashinfer import fp4_quantize, scaled_fp4_grouped_quantize
 from flashinfer.fused_moe import cutlass_fused_moe as flashinfer_cutlass_fused_moe
-from sgl_kernel import scaled_fp4_quant
+from sgl_kernel import scaled_fp4_quant, silu_and_mul
 from torch.nn import functional as F
 
 from sglang.srt.layers.moe.cutlass_moe import cutlass_moe_fp4
 from sglang.srt.layers.moe.cutlass_moe_params import CutlassMoEParams, CutlassMoEType
-from sglang.srt.layers.moe.flashinfer_cutedsl_moe import flashinfer_cutedsl_moe_masked
 from sglang.srt.layers.moe.topk import TopKConfig, select_experts
 
 if torch.cuda.get_device_capability() < (10, 0):
diff --git a/test/srt/test_function_call_parser.py b/test/srt/test_function_call_parser.py
index 2fcd28ea2..12f26982a 100644
--- a/test/srt/test_function_call_parser.py
+++ b/test/srt/test_function_call_parser.py
@@ -12,9 +12,6 @@ from sglang.srt.function_call.llama32_detector import Llama32Detector
 from sglang.srt.function_call.mistral_detector import MistralDetector
 from sglang.srt.function_call.pythonic_detector import PythonicDetector
 from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector
-from sglang.srt.function_call.qwen25_detector import Qwen25Detector
-from sglang.srt.utils.hf_transformers_utils import get_tokenizer
-from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
 
 
 class TestPythonicDetector(unittest.TestCase):
diff --git a/test/srt/test_fused_moe.py b/test/srt/test_fused_moe.py
index 57b2f944b..65d35c59a 100644
--- a/test/srt/test_fused_moe.py
+++ b/test/srt/test_fused_moe.py
@@ -1,7 +1,6 @@
 import unittest
 
 import torch
-import torch.nn.functional as F
 from tqdm import tqdm
 
 from sglang.srt.layers.activation import SiluAndMul
diff --git a/test/srt/test_hybrid_attn_backend.py b/test/srt/test_hybrid_attn_backend.py
index e43158a2e..a16203ae7 100644
--- a/test/srt/test_hybrid_attn_backend.py
+++ b/test/srt/test_hybrid_attn_backend.py
@@ -1,4 +1,3 @@
-import os
 import unittest
 from types import SimpleNamespace
 
diff --git a/test/srt/test_kv_events.py b/test/srt/test_kv_events.py
index d333738c7..601831d03 100644
--- a/test/srt/test_kv_events.py
+++ b/test/srt/test_kv_events.py
@@ -1,7 +1,6 @@
 import time
 import unittest
 
-import msgspec
 import requests
 import zmq
 from msgspec.msgpack import Decoder
@@ -10,8 +9,6 @@ from sglang.srt.disaggregation.kv_events import (
     AllBlocksCleared,
     BlockRemoved,
     BlockStored,
-    EventBatch,
-    KVCacheEvent,
     KVEventBatch,
 )
 from sglang.srt.utils import kill_process_tree
diff --git a/test/srt/test_load_weights_from_remote_instance.py b/test/srt/test_load_weights_from_remote_instance.py
index 71ab24d1d..bd722a920 100644
--- a/test/srt/test_load_weights_from_remote_instance.py
+++ b/test/srt/test_load_weights_from_remote_instance.py
@@ -21,7 +21,6 @@ import unittest
 import numpy as np
 import requests
 import torch
-import torch.distributed as dist
 import torch.multiprocessing as mp
 
 import sglang as sgl
diff --git a/test/srt/test_mamba_unittest.py b/test/srt/test_mamba_unittest.py
index 7bbca75e1..ab6b5436d 100644
--- a/test/srt/test_mamba_unittest.py
+++ b/test/srt/test_mamba_unittest.py
@@ -1,4 +1,3 @@
-import inspect
 import os
 import unittest
 
diff --git a/test/srt/test_modelopt_export.py b/test/srt/test_modelopt_export.py
index eb518f585..aa477ff67 100644
--- a/test/srt/test_modelopt_export.py
+++ b/test/srt/test_modelopt_export.py
@@ -7,7 +7,6 @@ and quantization workflow.
 
 import json
 import os
-import sys
 import tempfile
 import unittest
 from unittest.mock import Mock, patch
@@ -23,7 +22,7 @@ from sglang.srt.model_loader.loader import ModelOptModelLoader
 
 # Check if modelopt is available
 try:
-    import modelopt
+    import modelopt  # noqa: F401
 
     MODELOPT_AVAILABLE = True
 except ImportError:
diff --git a/test/srt/test_modelopt_loader.py b/test/srt/test_modelopt_loader.py
index f137318f0..a2bad70b5 100644
--- a/test/srt/test_modelopt_loader.py
+++ b/test/srt/test_modelopt_loader.py
@@ -5,8 +5,6 @@ This test module verifies the functionality of ModelOptModelLoader, which
 applies NVIDIA Model Optimizer quantization to models during loading.
 """
 
-import os
-import sys
 import unittest
 from unittest.mock import MagicMock, patch
 
diff --git a/test/srt/test_mscclpp.py b/test/srt/test_mscclpp.py
index 894598b3d..c30dbe00d 100644
--- a/test/srt/test_mscclpp.py
+++ b/test/srt/test_mscclpp.py
@@ -8,28 +8,20 @@ else
 fi
 """
 
-import itertools
 import os
 import random
 import socket
 import unittest
-from contextlib import contextmanager, nullcontext
-from typing import Any, List, Optional, Union
+from typing import Any
 
 import ray
 import torch
 import torch.distributed as dist
-from torch.distributed import ProcessGroup, ReduceOp
 
 from sglang.srt.distributed import init_distributed_environment
 from sglang.srt.distributed.communication_op import (  # noqa
     tensor_model_parallel_all_reduce,
 )
-from sglang.srt.distributed.device_communicators.custom_all_reduce import (
-    CustomAllreduce,
-)
-from sglang.srt.distributed.device_communicators.pymscclpp import PyMscclppCommunicator
-from sglang.srt.distributed.device_communicators.pynccl import PyNcclCommunicator
 from sglang.srt.distributed.parallel_state import (
     get_tensor_model_parallel_group,
     graph_capture,
@@ -37,7 +29,6 @@ from sglang.srt.distributed.parallel_state import (
     set_custom_all_reduce,
     set_mscclpp_all_reduce,
 )
-from sglang.srt.distributed.utils import StatelessProcessGroup
 from sglang.test.test_utils import CustomTestCase
 
 
diff --git a/test/srt/test_multi_tokenizer.py b/test/srt/test_multi_tokenizer.py
index 182454e5e..f705c061c 100644
--- a/test/srt/test_multi_tokenizer.py
+++ b/test/srt/test_multi_tokenizer.py
@@ -1,7 +1,6 @@
 import unittest
 from types import SimpleNamespace
 
-import sglang.srt.managers.io_struct as io_struct
 from sglang.srt.utils import kill_process_tree
 from sglang.test.run_eval import run_eval
 from sglang.test.test_utils import (
diff --git a/test/srt/test_original_logprobs.py b/test/srt/test_original_logprobs.py
index 1a50e3aa6..81ed0a403 100644
--- a/test/srt/test_original_logprobs.py
+++ b/test/srt/test_original_logprobs.py
@@ -17,7 +17,6 @@ import os
 import random
 import unittest
 
-import numpy as np
 import torch
 import torch.nn.functional as F
 from transformers import AutoModelForCausalLM, AutoTokenizer
diff --git a/test/srt/test_priority_scheduling.py b/test/srt/test_priority_scheduling.py
index befde130e..ef79188f8 100644
--- a/test/srt/test_priority_scheduling.py
+++ b/test/srt/test_priority_scheduling.py
@@ -2,7 +2,7 @@ import asyncio
 import os
 import re
 import unittest
-from typing import Any, Awaitable, Callable, List, Optional, Tuple
+from typing import Any, List, Optional, Tuple
 
 from sglang.srt.utils import kill_process_tree
 from sglang.test.test_utils import (
diff --git a/test/srt/test_quantization.py b/test/srt/test_quantization.py
index 990378d9b..3ce4a5ee9 100644
--- a/test/srt/test_quantization.py
+++ b/test/srt/test_quantization.py
@@ -1,8 +1,6 @@
 import json
-import os
 import unittest
 import warnings
-from datetime import datetime
 from types import SimpleNamespace
 
 from sglang.srt.utils import kill_process_tree
diff --git a/test/srt/test_schedule_policy.py b/test/srt/test_schedule_policy.py
index 0e33b6b25..2be092e31 100644
--- a/test/srt/test_schedule_policy.py
+++ b/test/srt/test_schedule_policy.py
@@ -6,7 +6,7 @@ from sglang.srt.managers.schedule_policy import (
     CacheAwarePolicy,
     SchedulePolicy,
 )
-from sglang.srt.mem_cache.radix_cache import RadixCache, TreeNode
+from sglang.srt.mem_cache.radix_cache import RadixCache
 from sglang.srt.sampling.sampling_params import SamplingParams
 from sglang.test.test_utils import CustomTestCase
 
diff --git a/test/srt/test_tokenizer_batch_encode.py b/test/srt/test_tokenizer_batch_encode.py
index 13d294d68..8d6e7539d 100644
--- a/test/srt/test_tokenizer_batch_encode.py
+++ b/test/srt/test_tokenizer_batch_encode.py
@@ -10,12 +10,10 @@ python3 -m unittest test_tokenizer_batch_encode.TestTokenizerBatchEncodeUnit.tes
 python3 -m unittest test_tokenizer_batch_encode.TestTokenizerBatchEncodeLogic.test_batch_processing_path
 """
 
-import asyncio
 import unittest
-from typing import List
-from unittest.mock import AsyncMock, Mock, call, patch
+from unittest.mock import Mock, patch
 
-from sglang.srt.managers.io_struct import GenerateReqInput, TokenizedGenerateReqInput
+from sglang.srt.managers.io_struct import GenerateReqInput
 from sglang.srt.managers.tokenizer_manager import TokenizerManager
 from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.test.test_utils import DEFAULT_SMALL_MODEL_NAME_FOR_TEST
diff --git a/test/srt/test_tokenizer_manager.py b/test/srt/test_tokenizer_manager.py
index d1817e6d9..04115fb99 100644
--- a/test/srt/test_tokenizer_manager.py
+++ b/test/srt/test_tokenizer_manager.py
@@ -12,7 +12,6 @@ python3 -m unittest test_tokenizer_manager.TestTokenizerManagerIntegration
 """
 
 import unittest
-from typing import List, Optional, Union
 from unittest.mock import Mock, patch
 
 from sglang.srt.managers.tokenizer_manager import TokenizerManager
diff --git a/test/srt/test_tracing.py b/test/srt/test_tracing.py
index 173b15b50..4e3763ac4 100644
--- a/test/srt/test_tracing.py
+++ b/test/srt/test_tracing.py
@@ -10,7 +10,6 @@ import requests
 import zmq
 
 from sglang import Engine
-from sglang.srt.managers.io_struct import TokenizedGenerateReqInput
 from sglang.srt.tracing.trace import *
 from sglang.srt.utils import get_zmq_socket, kill_process_tree
 from sglang.test.test_utils import (
diff --git a/test/srt/test_triton_fused_moe.py b/test/srt/test_triton_fused_moe.py
index f9b1ee884..d989494c9 100644
--- a/test/srt/test_triton_fused_moe.py
+++ b/test/srt/test_triton_fused_moe.py
@@ -1,7 +1,6 @@
 import unittest
 
 import torch
-import torch.nn.functional as F
 from tqdm import tqdm
 
 from sglang.srt.layers.activation import SiluAndMul
diff --git a/test/srt/test_vision_openai_server_a.py b/test/srt/test_vision_openai_server_a.py
index 2f727ecc1..7f8cc87b5 100644
--- a/test/srt/test_vision_openai_server_a.py
+++ b/test/srt/test_vision_openai_server_a.py
@@ -207,11 +207,12 @@ class TestDeepseekOCRServer(TestOpenAIMLLMServerBase):
 
 
 if __name__ == "__main__":
-    del (
-        TestOpenAIMLLMServerBase,
-        ImageOpenAITestMixin,
-        VideoOpenAITestMixin,
-        AudioOpenAITestMixin,
-        OmniOpenAITestMixin,
-    )
+    # Note: Cannot delete mixin classes imported via * since they're not in local scope
+    # del (
+    #     TestOpenAIMLLMServerBase,
+    #     ImageOpenAITestMixin,
+    #     VideoOpenAITestMixin,
+    #     AudioOpenAITestMixin,
+    #     OmniOpenAITestMixin,
+    # )
     unittest.main()
diff --git a/test/srt/test_vlm_input_format.py b/test/srt/test_vlm_input_format.py
index cc2ebcb3a..489f9cbba 100644
--- a/test/srt/test_vlm_input_format.py
+++ b/test/srt/test_vlm_input_format.py
@@ -7,7 +7,6 @@ import requests
 import torch
 from PIL import Image
 from transformers import (
-    AutoModel,
     AutoProcessor,
     Gemma3ForConditionalGeneration,
     Qwen2_5_VLForConditionalGeneration,
diff --git a/test/srt/xpu/test_intel_xpu_backend.py b/test/srt/xpu/test_intel_xpu_backend.py
index a0f301617..701769e75 100644
--- a/test/srt/xpu/test_intel_xpu_backend.py
+++ b/test/srt/xpu/test_intel_xpu_backend.py
@@ -3,7 +3,6 @@ Usage:
 python3 -m unittest test_intel_xpu_backend.TestIntelXPUBackend.test_latency_qwen_model
 """
 
-import os
 import unittest
 from functools import wraps