diff --git a/aidial_adapter_openai/audio_api/transcribe/adapter.py b/aidial_adapter_openai/audio_api/transcribe/adapter.py
index 44560f38..0a445486 100644
--- a/aidial_adapter_openai/audio_api/transcribe/adapter.py
+++ b/aidial_adapter_openai/audio_api/transcribe/adapter.py
@@ -43,7 +43,7 @@ def _get_usage(
 ) -> TokenUsage | None:
     # NOTE: whisper has completely different API for its responses
     duration: Any | None = getattr(chunk, "duration", None)
-    if duration is not None and isinstance(duration, (float, int)):
+    if duration is not None and isinstance(duration, float | int):
         return TokenUsage(prompt_tokens=int(duration))
 
     usage_dict: dict | None = getattr(chunk, "usage", None)  # type: ignore
diff --git a/aidial_adapter_openai/audio_api/transcribe/prompt.py b/aidial_adapter_openai/audio_api/transcribe/prompt.py
index 2a9f9e98..ede2858c 100644
--- a/aidial_adapter_openai/audio_api/transcribe/prompt.py
+++ b/aidial_adapter_openai/audio_api/transcribe/prompt.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import mimetypes
-from typing import Any, List
+from typing import Any
 
 from aidial_sdk.exceptions import RequestValidationError
 from pydantic import BaseModel
@@ -50,7 +50,7 @@ async def from_request(
 
         system_message = _collect_system_messages(messages)
 
-        audios: List[FileResource] = []
+        audios: list[FileResource] = []
 
         for message in result:
             for file in message.files:
diff --git a/aidial_adapter_openai/chat_completions/gpt.py b/aidial_adapter_openai/chat_completions/gpt.py
index 80503b54..2b4fae78 100644
--- a/aidial_adapter_openai/chat_completions/gpt.py
+++ b/aidial_adapter_openai/chat_completions/gpt.py
@@ -1,4 +1,4 @@
-from typing import AsyncIterator, Callable, Coroutine, List, Mapping, Tuple
+from collections.abc import AsyncIterator, Callable, Coroutine, Mapping
 
 from aidial_sdk.exceptions import InvalidRequestError
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
@@ -29,10 +29,10 @@
 
 async def multi_modal_truncate_prompt(
     request: dict,
-    messages: List[MultiModalMessage],
+    messages: list[MultiModalMessage],
     max_prompt_tokens: int,
     tokenizer: Tokenizer,
-) -> Tuple[List[MultiModalMessage], DiscardedMessages, TruncatedTokens]:
+) -> tuple[list[MultiModalMessage], DiscardedMessages, TruncatedTokens]:
     return await truncate_prompt(
         messages=messages,
         message_tokens=tokenizer.tokenize_request_message,
@@ -62,9 +62,9 @@ def _extract_max_prompt_tokens(request: dict) -> int | None:
 
 
 async def _truncate_messages(
-    request: dict, messages: List[MultiModalMessage], tokenizer: Tokenizer
-) -> Tuple[
-    List[MultiModalMessage],
+    request: dict, messages: list[MultiModalMessage], tokenizer: Tokenizer
+) -> tuple[
+    list[MultiModalMessage],
     DiscardedMessages | None,
     Callable[[], Coroutine[None, None, TruncatedTokens]],
 ]:
@@ -109,7 +109,7 @@ async def chat_completion(
     eliminate_empty_choices: bool,
 ) -> ResponseWithHeaders[AsyncIterator[dict] | dict]:
     n: int = request.get("n") or 1
-    messages: List[dict] = request["messages"]
+    messages: list[dict] = request["messages"]
     model_name = request["model"]
 
     multi_modal_messages = await ResourceProcessor(
diff --git a/aidial_adapter_openai/chat_completions/gpt_audio.py b/aidial_adapter_openai/chat_completions/gpt_audio.py
index f308959b..b3cc8e43 100644
--- a/aidial_adapter_openai/chat_completions/gpt_audio.py
+++ b/aidial_adapter_openai/chat_completions/gpt_audio.py
@@ -6,7 +6,8 @@
 DIAL-compatible format with attachments and stages.
 """
 
-from typing import AsyncIterator, TypeVar
+from collections.abc import AsyncIterator
+from typing import TypeVar
 
 from pydantic import BaseModel
 
diff --git a/aidial_adapter_openai/chat_completions/gpt_oss.py b/aidial_adapter_openai/chat_completions/gpt_oss.py
index 2c8e75bf..7c44a03e 100644
--- a/aidial_adapter_openai/chat_completions/gpt_oss.py
+++ b/aidial_adapter_openai/chat_completions/gpt_oss.py
@@ -18,7 +18,8 @@
 https://cookbook.openai.com/articles/gpt-oss/handle-raw-cot
 """
 
-from typing import AsyncIterator, Set, TypeVar
+from collections.abc import AsyncIterator
+from typing import TypeVar
 
 from pydantic import BaseModel
 
@@ -26,7 +27,7 @@
 
 
 class _ResponseTransformer(BaseModel):
-    opened_reasoning_stages: Set[int] = set()
+    opened_reasoning_stages: set[int] = set()
     """Indices of choices where a reasoning stage was open.
     """
 
diff --git a/aidial_adapter_openai/chat_completions/non_gpt.py b/aidial_adapter_openai/chat_completions/non_gpt.py
index 4a1fac3d..470c42f7 100644
--- a/aidial_adapter_openai/chat_completions/non_gpt.py
+++ b/aidial_adapter_openai/chat_completions/non_gpt.py
@@ -1,4 +1,4 @@
-from typing import AsyncIterator
+from collections.abc import AsyncIterator
 
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
 from openai.types.chat.chat_completion import ChatCompletion
diff --git a/aidial_adapter_openai/chat_completions/transformation.py b/aidial_adapter_openai/chat_completions/transformation.py
index 9625d078..d8c60e18 100644
--- a/aidial_adapter_openai/chat_completions/transformation.py
+++ b/aidial_adapter_openai/chat_completions/transformation.py
@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import List, Set, assert_never
+from typing import assert_never
 
 from aidial_sdk.exceptions import InvalidRequestError
 from openai.types.chat import (
@@ -46,15 +46,15 @@ class Error:
 
 class MessageTransformer:
     file_storage: FileStorage | None
-    errors: Set[Error]
-    images: List[ImageResource]
-    files: List[FileResource]
+    errors: set[Error]
+    images: list[ImageResource]
+    files: list[FileResource]
 
     def __init__(
         self,
         *,
         file_storage: FileStorage | None,
-        errors: Set[Error] | None = None,
+        errors: set[Error] | None = None,
     ):
         self.file_storage = file_storage
         self.errors = set() if errors is None else errors
@@ -81,12 +81,12 @@ async def try_download_resource(
             return None
 
     async def download_attachments(
-        self, attachments: List[dict]
-    ) -> List[ChatCompletionContentPartImageParam | File]:
+        self, attachments: list[dict]
+    ) -> list[ChatCompletionContentPartImageParam | File]:
         if attachments:
             logger.debug(f"original attachments: {attachments}")
 
-        ret: List[ChatCompletionContentPartImageParam | File] = []
+        ret: list[ChatCompletionContentPartImageParam | File] = []
         for attachment in attachments:
             if result := await self.download_attachment(attachment):
                 ret.append(result)
@@ -145,15 +145,15 @@ async def download_content(
         self,
         content: (
             str
-            | List[ChatCompletionContentPartParam | ContentArrayOfContentPart]
+            | list[ChatCompletionContentPartParam | ContentArrayOfContentPart]
         ),
-    ) -> List[ChatCompletionContentPartParam | ContentArrayOfContentPart]:
+    ) -> list[ChatCompletionContentPartParam | ContentArrayOfContentPart]:
         if isinstance(content, str):
             parts = [create_text_content_part(content)]
         else:
             parts = content
 
-        ret: List[
+        ret: list[
             ChatCompletionContentPartParam | ContentArrayOfContentPart
         ] = []
         for part in parts:
@@ -192,9 +192,9 @@ class ResourceProcessor(BaseModel):
     file_storage: FileStorage | None
 
     async def transform_messages(
-        self, messages: List[dict]
-    ) -> List[MultiModalMessage]:
-        errors: Set[Error] = set()
+        self, messages: list[dict]
+    ) -> list[MultiModalMessage]:
+        errors: set[Error] = set()
         transformations = [
             await MessageTransformer(
                 file_storage=self.file_storage, errors=errors
diff --git a/aidial_adapter_openai/completions.py b/aidial_adapter_openai/completions.py
index e78a8abe..03dd6083 100644
--- a/aidial_adapter_openai/completions.py
+++ b/aidial_adapter_openai/completions.py
@@ -1,4 +1,5 @@
-from typing import Any, AsyncIterator, Dict
+from collections.abc import AsyncIterator
+from typing import Any
 
 from aidial_sdk.exceptions import RequestValidationError
 from openai import AsyncAzureOpenAI, AsyncOpenAI, AsyncStream
@@ -18,7 +19,7 @@ def sanitize_text(text: str) -> str:
 
 def convert_to_chat_completions_response(
     chunk: Completion, is_stream: bool
-) -> Dict[str, Any]:
+) -> dict[str, Any]:
     converted_chunk = build_chunk(
         id=chunk.id,
         model=chunk.model,
@@ -37,7 +38,7 @@ def convert_to_chat_completions_response(
 
 async def chat_completion(
     *,
-    request: Dict[str, Any],
+    request: dict[str, Any],
     client: AsyncAzureOpenAI | AsyncOpenAI,
     prompt_template: str | None,
 ) -> AsyncIterator[dict] | dict:
diff --git a/aidial_adapter_openai/configuration/app_config.py b/aidial_adapter_openai/configuration/app_config.py
index e69ad6bf..f71f1cd3 100644
--- a/aidial_adapter_openai/configuration/app_config.py
+++ b/aidial_adapter_openai/configuration/app_config.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import os
-from typing import Dict, List, assert_never
+from typing import assert_never
 
 from aidial_sdk.exceptions import InternalServerError
 
@@ -39,23 +39,23 @@ class DeploymentAPIType(ExtraForbidModel):
 
 
 class ApplicationConfig(ExtraForbidModel):
-    TIKTOKEN_MODEL_MAPPING: Dict[str, str] = {}
+    TIKTOKEN_MODEL_MAPPING: dict[str, str] = {}
 
-    DALLE3_DEPLOYMENTS: List[str] = []
+    DALLE3_DEPLOYMENTS: list[str] = []
     DALLE3_AZURE_API_VERSION: str = "2024-02-01"
 
-    GPT_IMAGE_1_DEPLOYMENTS: List[str] = []
+    GPT_IMAGE_1_DEPLOYMENTS: list[str] = []
     GPT_IMAGE_1_AZURE_API_VERSION: str = "2025-04-01-preview"
 
-    MISTRAL_DEPLOYMENTS: List[str] = []
-    DATABRICKS_DEPLOYMENTS: List[str] = []
-    GPT4O_DEPLOYMENTS: List[str] = []
-    GPT4O_MINI_DEPLOYMENTS: List[str] = []
-    AZURE_AI_VISION_DEPLOYMENTS: List[str] = []
+    MISTRAL_DEPLOYMENTS: list[str] = []
+    DATABRICKS_DEPLOYMENTS: list[str] = []
+    GPT4O_DEPLOYMENTS: list[str] = []
+    GPT4O_MINI_DEPLOYMENTS: list[str] = []
+    AZURE_AI_VISION_DEPLOYMENTS: list[str] = []
 
-    API_VERSIONS_MAPPING: Dict[str, str] = {}
-    COMPLETION_DEPLOYMENTS_PROMPT_TEMPLATES: Dict[str, str] = {}
-    NON_STREAMING_DEPLOYMENTS: List[str] = []
+    API_VERSIONS_MAPPING: dict[str, str] = {}
+    COMPLETION_DEPLOYMENTS_PROMPT_TEMPLATES: dict[str, str] = {}
+    NON_STREAMING_DEPLOYMENTS: list[str] = []
     ELIMINATE_EMPTY_CHOICES: bool = False
 
     AUDIO_AZURE_API_VERSION: str = "2025-03-01-preview"
diff --git a/aidial_adapter_openai/dial_api/embedding_inputs.py b/aidial_adapter_openai/dial_api/embedding_inputs.py
index 6ffe3d6b..f77d8c05 100644
--- a/aidial_adapter_openai/dial_api/embedding_inputs.py
+++ b/aidial_adapter_openai/dial_api/embedding_inputs.py
@@ -1,8 +1,5 @@
+from collections.abc import AsyncIterator, Callable, Coroutine
 from typing import (
-    AsyncIterator,
-    Callable,
-    Coroutine,
-    List,
     TypeVar,
     assert_never,
     cast,
@@ -15,7 +12,7 @@
 _T = TypeVar("_T")
 
 _Coro = Coroutine[None, None, _T]
-_Tokens = List[int]
+_Tokens = list[int]
 
 
 async def reject_tokens(tokens: _Tokens):
@@ -25,7 +22,7 @@ async def reject_tokens(tokens: _Tokens):
     )
 
 
-async def reject_mixed(input: List[str | Attachment]):
+async def reject_mixed(input: list[str | Attachment]):
     raise RequestValidationError(
         "Embedding inputs composed of multiple texts and/or attachments aren't supported"
     )
@@ -37,7 +34,7 @@ async def collect_embedding_inputs(
     on_text: Callable[[str], _Coro[_T]],
     on_attachment: Callable[[Attachment], _Coro[_T]],
     on_tokens: Callable[[_Tokens], _Coro[_T]] = reject_tokens,
-    on_mixed: Callable[[List[str | Attachment]], _Coro[_T]] = reject_mixed,
+    on_mixed: Callable[[list[str | Attachment]], _Coro[_T]] = reject_mixed,
 ) -> AsyncIterator[_T]:
     async def _on_str_or_attachment(input: str | Attachment) -> _T:
         if isinstance(input, str):
@@ -70,7 +67,7 @@ async def _on_str_or_attachment(input: str | Attachment) -> _T:
         return
 
     for input in request.custom_input:
-        if isinstance(input, (str, Attachment)):
+        if isinstance(input, str | Attachment):
             yield await _on_str_or_attachment(input)
         elif isinstance(input, list):
             if len(input) == 0:
diff --git a/aidial_adapter_openai/dial_api/request.py b/aidial_adapter_openai/dial_api/request.py
index a39621c9..65e143ba 100644
--- a/aidial_adapter_openai/dial_api/request.py
+++ b/aidial_adapter_openai/dial_api/request.py
@@ -1,4 +1,4 @@
-from typing import Any, Type, TypeVar
+from typing import Any, TypeVar
 
 from aidial_sdk.exceptions import RequestValidationError
 from pydantic import BaseModel, ValidationError
@@ -6,7 +6,7 @@
 _T = TypeVar("_T", bound=BaseModel)
 
 
-def parse_configuration(cls: Type[_T], data: Any) -> _T | None:
+def parse_configuration(cls: type[_T], data: Any) -> _T | None:
     if (cf := data.get("custom_fields")) is None:
         return None
 
diff --git a/aidial_adapter_openai/dial_api/resource.py b/aidial_adapter_openai/dial_api/resource.py
index 1721e715..26a5fa8b 100644
--- a/aidial_adapter_openai/dial_api/resource.py
+++ b/aidial_adapter_openai/dial_api/resource.py
@@ -1,7 +1,6 @@
 import base64
 import mimetypes
 from abc import ABC, abstractmethod
-from typing import List
 
 from aidial_sdk.chat_completion import Attachment
 from pydantic import BaseModel, model_validator
@@ -25,9 +24,9 @@ class MissingContentTypeError(ValidationError):
 
 class UnsupportedContentTypeError(ValidationError):
     type: str
-    supported_types: List[str]
+    supported_types: list[str]
 
-    def __init__(self, *, message: str, type: str, supported_types: List[str]):
+    def __init__(self, *, message: str, type: str, supported_types: list[str]):
         self.type = type
         self.supported_types = supported_types
         super().__init__(message)
@@ -35,7 +34,7 @@ def __init__(self, *, message: str, type: str, supported_types: List[str]):
 
 class DialResource(ABC, BaseModel):
     entity_name: str | None = None
-    supported_types: List[str] | None = None
+    supported_types: list[str] | None = None
 
     @abstractmethod
     async def download(self, storage: FileStorage | None) -> Resource: ...
diff --git a/aidial_adapter_openai/dial_api/sdk_adapter.py b/aidial_adapter_openai/dial_api/sdk_adapter.py
index 9b25cd16..aa924455 100644
--- a/aidial_adapter_openai/dial_api/sdk_adapter.py
+++ b/aidial_adapter_openai/dial_api/sdk_adapter.py
@@ -1,4 +1,4 @@
-from typing import Callable, Coroutine
+from collections.abc import Callable, Coroutine
 
 import fastapi
 from aidial_sdk.chat_completion import Request as DIALRequest
diff --git a/aidial_adapter_openai/dial_api/storage.py b/aidial_adapter_openai/dial_api/storage.py
index 886527cc..d498020d 100644
--- a/aidial_adapter_openai/dial_api/storage.py
+++ b/aidial_adapter_openai/dial_api/storage.py
@@ -1,7 +1,7 @@
 import base64
 import hashlib
 import mimetypes
-from typing import Mapping, Optional
+from collections.abc import Mapping
 from urllib.parse import unquote, urljoin
 
 from pydantic import BaseModel, SecretStr
@@ -28,7 +28,7 @@ class FileStorage(BaseModel):
     dial_url: str
     api_key: SecretStr
 
-    bucket: Optional[Bucket] = None
+    bucket: Bucket | None = None
 
     @property
     def headers(self) -> Mapping[str, str]:
@@ -129,14 +129,14 @@ def _compute_hash_digest(file_content: str | bytes) -> str:
 
 DIAL_USE_FILE_STORAGE = get_env_bool("DIAL_USE_FILE_STORAGE", False)
 
-DIAL_URL: Optional[str] = None
+DIAL_URL: str | None = None
 if DIAL_USE_FILE_STORAGE:
     DIAL_URL = get_env(
         "DIAL_URL", "DIAL_URL must be set to use the DIAL file storage"
     )
 
 
-def create_file_storage(headers: Mapping[str, str]) -> Optional[FileStorage]:
+def create_file_storage(headers: Mapping[str, str]) -> FileStorage | None:
     if not DIAL_USE_FILE_STORAGE or DIAL_URL is None:
         return None
 
diff --git a/aidial_adapter_openai/embeddings/azure_ai_vision.py b/aidial_adapter_openai/embeddings/azure_ai_vision.py
index 65348434..f35fc09f 100644
--- a/aidial_adapter_openai/embeddings/azure_ai_vision.py
+++ b/aidial_adapter_openai/embeddings/azure_ai_vision.py
@@ -31,7 +31,8 @@
 """
 
 import asyncio
-from typing import AsyncIterator, List, assert_never
+from collections.abc import AsyncIterator
+from typing import assert_never
 
 import httpx
 from aidial_sdk.chat_completion.request import Attachment
@@ -69,7 +70,7 @@ def _get_auth_headers(creds: OpenAICreds) -> dict[str, str]:
 
 
 class VectorizeResponse(ExtraAllowedModel):
-    vector: List[float]
+    vector: list[float]
 
 
 async def embeddings(
@@ -95,7 +96,7 @@ async def on_attachment(attachment: Attachment) -> Resource:
         on_attachment=on_attachment,
     )
 
-    inputs: List[str | Resource] = [input async for input in inputs_iter]
+    inputs: list[str | Resource] = [input async for input in inputs_iter]
 
     headers = _get_auth_headers(creds)
 
diff --git a/aidial_adapter_openai/endpoints/chat_completion.py b/aidial_adapter_openai/endpoints/chat_completion.py
index 193469e3..b9af390b 100644
--- a/aidial_adapter_openai/endpoints/chat_completion.py
+++ b/aidial_adapter_openai/endpoints/chat_completion.py
@@ -1,4 +1,5 @@
-from typing import Mapping, assert_never
+from collections.abc import Mapping
+from typing import assert_never
 
 import fastapi
 from fastapi import Request
diff --git a/aidial_adapter_openai/exception_handlers.py b/aidial_adapter_openai/exception_handlers.py
index a582fb91..bcbc9d4a 100644
--- a/aidial_adapter_openai/exception_handlers.py
+++ b/aidial_adapter_openai/exception_handlers.py
@@ -24,7 +24,7 @@ def to_adapter_exception(exc: Exception) -> AdapterException:
 
 
 def _convert_to_adapter_exception(exc: Exception) -> AdapterException:
-    if isinstance(exc, (DialException, ResponseWrapper)):
+    if isinstance(exc, DialException | ResponseWrapper):
         return exc
 
     if isinstance(exc, httpx.HTTPStatusError):
diff --git a/aidial_adapter_openai/image_generation/prompt.py b/aidial_adapter_openai/image_generation/prompt.py
index 06516dbb..b6f7a565 100644
--- a/aidial_adapter_openai/image_generation/prompt.py
+++ b/aidial_adapter_openai/image_generation/prompt.py
@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import Any, List
+from typing import Any
 
 from aidial_sdk.exceptions import InvalidRequestError
 from pydantic import BaseModel
@@ -15,7 +15,7 @@
 
 class ImageGenPrompt(BaseModel):
     text_prompt: str
-    images: List[Resource]
+    images: list[Resource]
 
     @classmethod
     async def from_request(
@@ -26,7 +26,7 @@ async def from_request(
         ).transform_messages(data["messages"])
 
         text_prompt = ""
-        images: List[Resource] = []
+        images: list[Resource] = []
 
         for message in result:
             text_prompt += collect_message_text_content(message.raw_message)
diff --git a/aidial_adapter_openai/responses/adapter.py b/aidial_adapter_openai/responses/adapter.py
index 1807161c..88bd334e 100644
--- a/aidial_adapter_openai/responses/adapter.py
+++ b/aidial_adapter_openai/responses/adapter.py
@@ -1,6 +1,7 @@
 import json
 import logging
-from typing import Any, AsyncIterator, Dict, List
+from collections.abc import AsyncIterator
+from typing import Any
 
 from aidial_sdk.exceptions import RequestValidationError
 from openai import (
@@ -36,15 +37,15 @@
 )
 
 
-def _validate_request(request: Dict[str, Any]) -> None:
-    errors: List[str] = []
+def _validate_request(request: dict[str, Any]) -> None:
+    errors: list[str] = []
 
     if (n := request.get("n")) not in [None, 1]:
         errors.append(
             f"The deployment doesn't support request.n parameter other than 1, but got {n}."
         )
 
-    unsupported_params: List[str] = []
+    unsupported_params: list[str] = []
     for param in [
         "stop",
         "seed",
@@ -88,7 +89,7 @@ class ResponsesConfig(ExtraAllowedModel):
     )
 
 
-def _get_configuration(request: Dict[str, Any]) -> ResponsesConfig:
+def _get_configuration(request: dict[str, Any]) -> ResponsesConfig:
     configuration = (
         parse_configuration(ResponsesConfig, request) or ResponsesConfig()
     )
@@ -104,7 +105,7 @@ def _get_configuration(request: Dict[str, Any]) -> ResponsesConfig:
 
 async def chat_completion(
     *,
-    request: Dict[str, Any],
+    request: dict[str, Any],
     client: AsyncAzureOpenAI | AsyncOpenAI,
     file_storage: FileStorage | None,
 ) -> AsyncIterator[dict] | dict:
diff --git a/aidial_adapter_openai/responses/converter.py b/aidial_adapter_openai/responses/converter.py
index 767ebced..301de924 100644
--- a/aidial_adapter_openai/responses/converter.py
+++ b/aidial_adapter_openai/responses/converter.py
@@ -1,4 +1,5 @@
-from typing import Generator, List, assert_never
+from collections.abc import Generator
+from typing import assert_never
 
 from aidial_sdk.chat_completion.request import CustomContent, Stage, Status
 from aidial_sdk.exceptions import RequestValidationError
@@ -118,7 +119,7 @@ def convert_tool_choice(
     assert_never(tool_choice)
 
 
-def convert_tools(tools: List[ChatCompletionToolParam]) -> List[ToolParam]:
+def convert_tools(tools: list[ChatCompletionToolParam]) -> list[ToolParam]:
     def _convert_tool(tool: ChatCompletionToolParam) -> ToolParam:
         function = tool["function"]
         return FunctionToolParam(
@@ -264,17 +265,17 @@ def _convert_message(
 
 
 def convert_messages(
-    messages: List[ChatCompletionMessageParam],
+    messages: list[ChatCompletionMessageParam],
 ) -> ResponseInputParam:
     return [
         param for message in messages for param in _convert_message(message)
     ]
 
 
-def _convert_output(output: List[ResponseOutputItem]) -> ChatCompletionMessage:
+def _convert_output(output: list[ResponseOutputItem]) -> ChatCompletionMessage:
     text_content = ""
-    annotations: List[Annotation] = []
-    tool_calls: List[ChatCompletionMessageToolCall] = []
+    annotations: list[Annotation] = []
+    tool_calls: list[ChatCompletionMessageToolCall] = []
     custom_content: CustomContent | None = None
 
     for item in output:
@@ -307,7 +308,7 @@ def _convert_output(output: List[ResponseOutputItem]) -> ChatCompletionMessage:
 
             case ResponseReasoningItem(summary=summary):
                 if summary:
-                    stages: List[Stage] = []
+                    stages: list[Stage] = []
                     for index, summary_part in enumerate(summary):
                         suffix = "" if index == 0 else f" #{index + 1}"
                         stages.append(
diff --git a/aidial_adapter_openai/responses/event_handler.py b/aidial_adapter_openai/responses/event_handler.py
index 41e8cee4..26a06a16 100644
--- a/aidial_adapter_openai/responses/event_handler.py
+++ b/aidial_adapter_openai/responses/event_handler.py
@@ -1,6 +1,7 @@
 import json
 import logging
-from typing import Dict, Generator, assert_never
+from collections.abc import Generator
+from typing import assert_never
 
 import openai
 import pydantic
@@ -117,7 +118,7 @@ class EventHandler(pydantic.BaseModel):
     created_: int | None = None
     model_: str | None = None
 
-    tool_calls: Dict[str, int] = {}
+    tool_calls: dict[str, int] = {}
     """Map item_id for a tool call onto its index in the chat completion response
     """
 
diff --git a/aidial_adapter_openai/utils/adapter_exception.py b/aidial_adapter_openai/utils/adapter_exception.py
index 823a1122..d5024639 100644
--- a/aidial_adapter_openai/utils/adapter_exception.py
+++ b/aidial_adapter_openai/utils/adapter_exception.py
@@ -1,5 +1,6 @@
 import json
-from typing import Any, MutableMapping
+from collections.abc import MutableMapping
+from typing import Any
 
 from aidial_sdk.exceptions import HTTPException as DialException
 from fastapi.responses import Response as FastAPIResponse
@@ -25,11 +26,7 @@ def __init__(
     def __repr__(self):
         # headers field is omitted deliberately
         # since it may contain sensitive information
-        return "%s(content=%r, status_code=%r)" % (
-            self.__class__.__name__,
-            self.content,
-            self.status_code,
-        )
+        return f"{self.__class__.__name__}(content={self.content!r}, status_code={self.status_code!r})"
 
     def to_fastapi_response(self) -> FastAPIResponse:
         return FastAPIResponse(
diff --git a/aidial_adapter_openai/utils/auth.py b/aidial_adapter_openai/utils/auth.py
index 6815845d..5ddbf767 100644
--- a/aidial_adapter_openai/utils/auth.py
+++ b/aidial_adapter_openai/utils/auth.py
@@ -1,6 +1,6 @@
 import os
 import time
-from typing import Mapping
+from collections.abc import Mapping
 
 from aidial_sdk.exceptions import HTTPException as DialException
 from azure.core.credentials import AccessToken
diff --git a/aidial_adapter_openai/utils/caching.py b/aidial_adapter_openai/utils/caching.py
index b3a7c8ee..76ad7455 100644
--- a/aidial_adapter_openai/utils/caching.py
+++ b/aidial_adapter_openai/utils/caching.py
@@ -1,5 +1,6 @@
 import time
-from typing import Any, Callable, Coroutine, Mapping
+from collections.abc import Callable, Coroutine, Mapping
+from typing import Any
 
 _DIAL_CACHE_BREAKPOINT_PATH = "X-DIAL-CACHE-BREAKPOINT-PATH"
 _DIAL_CACHE_EXPIRE_AT = "X-DIAL-CACHE-EXPIRE-AT"
diff --git a/aidial_adapter_openai/utils/chat_completion_response.py b/aidial_adapter_openai/utils/chat_completion_response.py
index 968cfa35..1822634d 100644
--- a/aidial_adapter_openai/utils/chat_completion_response.py
+++ b/aidial_adapter_openai/utils/chat_completion_response.py
@@ -1,4 +1,5 @@
-from typing import Any, Iterable, Literal, Self, Set
+from collections.abc import Iterable
+from typing import Any, Literal, Self
 
 from aidial_sdk.utils.merge_chunks import merge_chat_completion_chunks
 from pydantic import BaseModel
@@ -26,7 +27,7 @@ def messages(self) -> Iterable[Any]:
     def has_messages(self) -> bool:
         return len(list(self.messages)) > 0
 
-    def get_missing_finish_reasons(self, n: int) -> Set[int]:
+    def get_missing_finish_reasons(self, n: int) -> set[int]:
         missing = set(range(n))
         for choice in self.response.get("choices") or []:
             if choice.get("finish_reason") is not None:
diff --git a/aidial_adapter_openai/utils/concurrency.py b/aidial_adapter_openai/utils/concurrency.py
index aba88f8f..891dd0a3 100644
--- a/aidial_adapter_openai/utils/concurrency.py
+++ b/aidial_adapter_openai/utils/concurrency.py
@@ -1,7 +1,8 @@
 import asyncio
 import os
+from collections.abc import Callable
 from concurrent.futures import ThreadPoolExecutor
-from typing import Callable, TypeVar
+from typing import TypeVar
 
 _THREAD_POOL_SIZE = os.getenv("THREAD_POOL_SIZE")
 _THREAD_POOL_SIZE = (
diff --git a/aidial_adapter_openai/utils/env.py b/aidial_adapter_openai/utils/env.py
index 4a2146d9..9c929973 100644
--- a/aidial_adapter_openai/utils/env.py
+++ b/aidial_adapter_openai/utils/env.py
@@ -1,11 +1,12 @@
 import json
 import os
-from typing import Callable, Dict, List, Optional, TypeVar
+from collections.abc import Callable
+from typing import TypeVar
 
 from aidial_adapter_openai.utils.log_config import logger
 
 
-def get_env(name: str, err_msg: Optional[str] = None) -> str:
+def get_env(name: str, err_msg: str | None = None) -> str:
     if (val := os.getenv(name)) is not None:
         return val
     raise Exception(err_msg or f"{name} env variable is not set")
@@ -17,13 +18,13 @@ def get_env_bool(name: str, default: bool = False) -> bool:
     return default
 
 
-def get_env_list(name: str) -> List[str] | None:
+def get_env_list(name: str) -> list[str] | None:
     if (value := os.getenv(name)) is not None:
         return list(map(str.strip, (value).split(",")))
     return None
 
 
-def get_env_dict(key: str) -> Dict[str, str] | None:
+def get_env_dict(key: str) -> dict[str, str] | None:
     if (value := os.getenv(key)) is not None:
         try:
             return json.loads(value)
@@ -41,7 +42,7 @@ def get_env_var(
     parser: Callable[[str], _T],
     name: str,
     *,
-    deprecated_names: List[str] | None = None,
+    deprecated_names: list[str] | None = None,
 ) -> _T:
     for alt in deprecated_names or []:
         if os.getenv(alt) is not None:
diff --git a/aidial_adapter_openai/utils/httpx.py b/aidial_adapter_openai/utils/httpx.py
index 61396c8a..4cd32bb1 100644
--- a/aidial_adapter_openai/utils/httpx.py
+++ b/aidial_adapter_openai/utils/httpx.py
@@ -1,6 +1,6 @@
 import logging
 import time
-from typing import Any, Tuple
+from typing import Any
 
 import httpx
 
@@ -20,7 +20,7 @@ def _elapsed_ms(start_time: float, last_time: float | None = None) -> int:
 TraceCallback = Any  # `httpcore` doesn’t publish a public protocol for this
 
 
-def _build_trace() -> Tuple[TraceCtx, TraceCallback]:
+def _build_trace() -> tuple[TraceCtx, TraceCallback]:
     trace_ctx: TraceCtx = {}
     starts: dict[str, float] = {}
 
diff --git a/aidial_adapter_openai/utils/multi_modal_message.py b/aidial_adapter_openai/utils/multi_modal_message.py
index 62e1967f..e57b1e65 100644
--- a/aidial_adapter_openai/utils/multi_modal_message.py
+++ b/aidial_adapter_openai/utils/multi_modal_message.py
@@ -1,5 +1,3 @@
-from typing import List
-
 from openai.types.chat import ChatCompletionContentPartTextParam
 from openai.types.chat.chat_completion_content_part_param import File
 from pydantic import BaseModel
@@ -24,6 +22,6 @@ def create_text_content_part(text: str) -> ChatCompletionContentPartTextParam:
 
 
 class MultiModalMessage(BaseModel):
-    images: List[ImageResource] = []
-    files: List[FileResource] = []
+    images: list[ImageResource] = []
+    files: list[FileResource] = []
     raw_message: dict
diff --git a/aidial_adapter_openai/utils/parsers.py b/aidial_adapter_openai/utils/parsers.py
index eb640c61..9e1b1234 100644
--- a/aidial_adapter_openai/utils/parsers.py
+++ b/aidial_adapter_openai/utils/parsers.py
@@ -1,7 +1,7 @@
 import re
 from http import HTTPStatus
 from json import JSONDecodeError
-from typing import Any, Dict
+from typing import Any
 
 from aidial_sdk.exceptions import HTTPException, InvalidRequestError
 from fastapi import Request
@@ -122,7 +122,7 @@ def try_parse(
 azure_video_api_parser = EndpointParser(name="video/generations")
 
 
-async def parse_body(request: Request) -> Dict[str, Any]:
+async def parse_body(request: Request) -> dict[str, Any]:
     try:
         data = await request.json()
     except JSONDecodeError as e:
diff --git a/aidial_adapter_openai/utils/reflection.py b/aidial_adapter_openai/utils/reflection.py
index effb3f8d..9bddd068 100644
--- a/aidial_adapter_openai/utils/reflection.py
+++ b/aidial_adapter_openai/utils/reflection.py
@@ -1,6 +1,7 @@
 import functools
 import inspect
-from typing import Any, Callable, Coroutine, TypeVar
+from collections.abc import Callable, Coroutine
+from typing import Any, TypeVar
 
 from aidial_sdk.exceptions import InvalidRequestError
 
diff --git a/aidial_adapter_openai/utils/resource/base.py b/aidial_adapter_openai/utils/resource/base.py
index caef5a2d..5ace5b6f 100644
--- a/aidial_adapter_openai/utils/resource/base.py
+++ b/aidial_adapter_openai/utils/resource/base.py
@@ -41,7 +41,7 @@ def to_data_url(self) -> str:
         return f"{self._to_data_url_prefix(self.type)}{self.data_base64}"
 
     @staticmethod
-    def parse_data_url_content_type(data_url: str) -> Optional[str]:
+    def parse_data_url_content_type(data_url: str) -> str | None:
         pattern = r"^data:([^;]+);base64,"
         match = re.match(pattern, data_url)
         return None if match is None else match.group(1)
diff --git a/aidial_adapter_openai/utils/resource/image.py b/aidial_adapter_openai/utils/resource/image.py
index b34d6d7c..d6f3c200 100644
--- a/aidial_adapter_openai/utils/resource/image.py
+++ b/aidial_adapter_openai/utils/resource/image.py
@@ -1,5 +1,5 @@
 from io import BytesIO
-from typing import Literal, Optional, assert_never
+from typing import Literal, assert_never
 
 from openai.types.chat import ChatCompletionContentPartImageParam
 from PIL import Image
@@ -39,7 +39,7 @@ class ImageResource(BaseModel):
 
     @classmethod
     def _from_resource(
-        cls, image: Resource, detail: Optional[ImageDetail]
+        cls, image: Resource, detail: ImageDetail | None
     ) -> "ImageResource":
         with Image.open(BytesIO(image.data)) as img:
             width, height = img.size
@@ -53,7 +53,7 @@ def _from_resource(
 
     @classmethod
     async def from_resource(
-        cls, image: Resource, detail: Optional[ImageDetail]
+        cls, image: Resource, detail: ImageDetail | None
     ) -> "ImageResource":
         return await run_in_threadpool(
             lambda: cls._from_resource(image, detail)
diff --git a/aidial_adapter_openai/utils/sse_stream.py b/aidial_adapter_openai/utils/sse_stream.py
index 3090630e..e52bef23 100644
--- a/aidial_adapter_openai/utils/sse_stream.py
+++ b/aidial_adapter_openai/utils/sse_stream.py
@@ -1,5 +1,6 @@
 import json
-from typing import Any, AsyncIterator, Mapping
+from collections.abc import AsyncIterator, Mapping
+from typing import Any
 
 from aidial_adapter_openai.utils.adapter_exception import AdapterException
 
diff --git a/aidial_adapter_openai/utils/streaming.py b/aidial_adapter_openai/utils/streaming.py
index f682e4fe..f72dd3ae 100644
--- a/aidial_adapter_openai/utils/streaming.py
+++ b/aidial_adapter_openai/utils/streaming.py
@@ -1,16 +1,9 @@
 import logging
+from collections.abc import AsyncIterator, Callable, Coroutine, Generator
 from dataclasses import dataclass
 from time import time
 from typing import (
-    AsyncIterator,
-    Callable,
-    Coroutine,
-    Generator,
     Generic,
-    List,
-    Optional,
-    Set,
-    Tuple,
     TypeVar,
     assert_never,
 )
@@ -46,8 +39,8 @@ def build_chunk(
     *,
     id: str,
     model: str,
-    finish_reason: Optional[str],
-    message: dict | List[dict],
+    finish_reason: str | None,
+    message: dict | list[dict],
     created: int,
     is_stream: bool,
     **extra,
@@ -81,7 +74,7 @@ async def generate_stream(
         [ChatCompletionResponse], Coroutine[None, None, int]
     ],
     model: str,
-    discarded_messages: Optional[list[int]],
+    discarded_messages: list[int] | None,
     eliminate_empty_choices: bool,
 ) -> AsyncIterator[dict]:
     empty_chunk = build_chunk(
@@ -117,7 +110,7 @@ async def set_usage(
 
     def set_default_finish_reasons(
         chunk: dict | None,
-        missing_indices: Set[int],
+        missing_indices: set[int],
         default_finish_reason: str,
     ) -> dict:
         def _set_reason(choice: dict) -> dict:
@@ -220,7 +213,7 @@ def block_response_to_streaming_chunk(response: dict) -> dict:
     return response
 
 
-def streaming_chunks_to_block_response(chunks: List[dict]) -> dict:
+def streaming_chunks_to_block_response(chunks: list[dict]) -> dict:
     response = merge_chat_completion_chunks(*chunks)
     response["object"] = "chat.completion"
 
@@ -306,7 +299,7 @@ async def map_stream_generator(
 
 
 async def map_stream(
-    func: Callable[[T], Optional[V]], iterator: AsyncIterator[T]
+    func: Callable[[T], V | None], iterator: AsyncIterator[T]
 ) -> AsyncIterator[V]:
     async for item in iterator:
         new_item = func(item)
@@ -325,7 +318,7 @@ async def prepend(
 
 async def peek_head(
     iterator: AsyncIterator[T],
-) -> Tuple[T | None, AsyncIterator[T]]:
+) -> tuple[T | None, AsyncIterator[T]]:
     try:
         val = await iterator.__anext__()
         return val, iterator
diff --git a/aidial_adapter_openai/utils/tokenizer.py b/aidial_adapter_openai/utils/tokenizer.py
index ae271430..ed01cb8e 100644
--- a/aidial_adapter_openai/utils/tokenizer.py
+++ b/aidial_adapter_openai/utils/tokenizer.py
@@ -4,8 +4,9 @@
 
 import json
 from abc import ABC, abstractmethod
+from collections.abc import Callable, Coroutine
 from functools import cached_property
-from typing import Any, Callable, Coroutine, Generic, List, Set, TypeVar
+from typing import Any, Generic, TypeVar
 
 from aidial_sdk.exceptions import InternalServerError
 from tiktoken import Encoding, encoding_for_model
@@ -118,7 +119,7 @@ def _tokens_per_request_message_name(self) -> int:
         return 1
 
     async def tokenize_request(
-        self, original_request: dict, messages: List[MessageType]
+        self, original_request: dict, messages: list[MessageType]
     ) -> int:
         tokens = self.TOKENS_PER_REQUEST
 
@@ -188,7 +189,7 @@ async def _tokenize_message(
 
 class Tokenizer(BaseTokenizer[MultiModalMessage]):
     image_tokenizer: ImageTokenizer | None
-    warnings: Set[str]
+    warnings: set[str]
 
     def __init__(
         self, *, model: str, image_tokenizer: ImageTokenizer | None = None
@@ -237,7 +238,7 @@ async def tokenize_request_message(self, message: MultiModalMessage) -> int:
         return tokens
 
     async def tokenize_request(
-        self, original_request: dict, messages: List[MultiModalMessage]
+        self, original_request: dict, messages: list[MultiModalMessage]
     ) -> int:
         tokens = await super().tokenize_request(original_request, messages)
 
diff --git a/aidial_adapter_openai/utils/truncate_prompt.py b/aidial_adapter_openai/utils/truncate_prompt.py
index e4e5178c..5ae25a0b 100644
--- a/aidial_adapter_openai/utils/truncate_prompt.py
+++ b/aidial_adapter_openai/utils/truncate_prompt.py
@@ -1,4 +1,5 @@
-from typing import Callable, Coroutine, List, Set, Tuple, TypeVar
+from collections.abc import Callable, Coroutine
+from typing import TypeVar
 
 from aidial_sdk.exceptions import (
     TruncatePromptSystemAndLastUserError,
@@ -7,21 +8,21 @@
 
 _T = TypeVar("_T")
 
-DiscardedMessages = List[int]
+DiscardedMessages = list[int]
 TruncatedTokens = int
 
 
 async def truncate_prompt(
-    messages: List[_T],
+    messages: list[_T],
     message_tokens: Callable[[_T], Coroutine[None, None, int]],
     is_system_message: Callable[[_T], bool],
     max_prompt_tokens: int,
     initial_prompt_tokens: int,
-) -> Tuple[List[_T], DiscardedMessages, TruncatedTokens]:
+) -> tuple[list[_T], DiscardedMessages, TruncatedTokens]:
     prompt_tokens = initial_prompt_tokens
 
     system_messages_count = 0
-    kept_messages: Set[int] = set()
+    kept_messages: set[int] = set()
 
     # Count system messages first
     for idx, message_holder in enumerate(messages):
diff --git a/aidial_adapter_openai/utils/validation.py b/aidial_adapter_openai/utils/validation.py
index ea40724b..2bffce93 100644
--- a/aidial_adapter_openai/utils/validation.py
+++ b/aidial_adapter_openai/utils/validation.py
@@ -1,10 +1,10 @@
-from typing import Any, Tuple, Type, Union
+from typing import Any
 
 from aidial_sdk.exceptions import InvalidRequestError
 
 
 def _ensure_type(
-    name: str, value: Any, expected: Union[Type, Tuple[Type, ...]]
+    name: str, value: Any, expected: type | tuple[type, ...]
 ) -> Any:
     expected = expected if isinstance(expected, tuple) else (expected,)
     if isinstance(value, expected):
diff --git a/aidial_adapter_openai/video_generation/azure/adapter.py b/aidial_adapter_openai/video_generation/azure/adapter.py
index 5b998883..28e34595 100644
--- a/aidial_adapter_openai/video_generation/azure/adapter.py
+++ b/aidial_adapter_openai/video_generation/azure/adapter.py
@@ -1,5 +1,5 @@
 import asyncio
-from typing import Any, Dict, List, assert_never
+from typing import Any, assert_never
 
 import fastapi
 from aidial_sdk.chat_completion import Choice, Stage
@@ -29,15 +29,15 @@
 )
 
 
-def _validate_request(request: Dict[str, Any]) -> None:
-    errors: List[str] = []
+def _validate_request(request: dict[str, Any]) -> None:
+    errors: list[str] = []
 
     if (n := request.get("n")) not in [None, 1]:
         errors.append(
             f"The deployment doesn't support request.n parameter other than 1, but got {n}."
         )
 
-    unsupported_params: List[str] = []
+    unsupported_params: list[str] = []
     for param in [
         "stop",
         "seed",
@@ -95,7 +95,7 @@ async def _poll_job(
     client: AzureVideoAPIClient,
     job_id: str,
     polling_interval: float,
-) -> List[VideoGeneration]:
+) -> list[VideoGeneration]:
     while True:
         await asyncio.sleep(polling_interval)
 
@@ -138,7 +138,7 @@ async def _download_videos(
     response: DIALResponse,
     choice: Choice,
     client: AzureVideoAPIClient,
-    video_generations: List[VideoGeneration],
+    video_generations: list[VideoGeneration],
     storage: FileStorage | None,
 ):
     n = len(video_generations)
@@ -167,7 +167,7 @@ async def _download_videos(
 async def chat_completion(
     *,
     request: fastapi.Request,
-    request_body: Dict[str, Any],
+    request_body: dict[str, Any],
     creds: OpenAICreds,
     deployment_id: str,
     upstream_endpoint: str,
diff --git a/aidial_adapter_openai/video_generation/azure/client.py b/aidial_adapter_openai/video_generation/azure/client.py
index d1b78e61..989ea675 100644
--- a/aidial_adapter_openai/video_generation/azure/client.py
+++ b/aidial_adapter_openai/video_generation/azure/client.py
@@ -1,5 +1,5 @@
 import json
-from typing import Dict, List, Literal, NoReturn, Self
+from typing import Literal, NoReturn, Self
 
 import httpx
 from aidial_sdk.exceptions import InternalServerError, InvalidRequestError
@@ -23,7 +23,7 @@ class VideoGenerationJob(BaseModel):
 
     id: str
     status: JobStatus
-    generations: List[VideoGeneration] | None = None
+    generations: list[VideoGeneration] | None = None
     failure_reason: (
         str | Literal["input_moderation", "internal_error"] | None
     ) = None
@@ -57,7 +57,7 @@ def _client(self) -> httpx.AsyncClient:
         return get_http_client()
 
     @property
-    def _headers(self) -> Dict[str, str]:
+    def _headers(self) -> dict[str, str]:
         headers = {"Content-Type": "application/json"}
         if key := self.creds.get("api_key"):
             headers["api-key"] = key
@@ -66,7 +66,7 @@ def _headers(self) -> Dict[str, str]:
         return headers
 
     @property
-    def _params(self) -> Dict[str, str]:
+    def _params(self) -> dict[str, str]:
         return {"api-version": "preview"}
 
     @property
diff --git a/aidial_adapter_openai/video_generation/azure/prompt.py b/aidial_adapter_openai/video_generation/azure/prompt.py
index 86a59953..54d719db 100644
--- a/aidial_adapter_openai/video_generation/azure/prompt.py
+++ b/aidial_adapter_openai/video_generation/azure/prompt.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 
 import mimetypes
-from typing import Any, List, Tuple
+from typing import Any
 
 from aidial_sdk.exceptions import InvalidRequestError
 from httpx._types import RequestFiles
@@ -21,7 +21,7 @@
 
 class VideoGenPrompt(BaseModel):
     prompt: str
-    resources: List[Resource]
+    resources: list[Resource]
 
     @classmethod
     async def from_request(
@@ -40,7 +40,7 @@ async def from_request(
             ).transform_messages([last_message])
         )[0]
 
-        resources: List[Resource] = []
+        resources: list[Resource] = []
         for image in multi_modal_message.images:
             resources.append(image.image)
 
@@ -49,7 +49,7 @@ async def from_request(
 
         return cls(prompt=prompt, resources=resources)
 
-    def get_files(self) -> Tuple[List[InpaintItem], RequestFiles]:
+    def get_files(self) -> tuple[list[InpaintItem], RequestFiles]:
         items, files = [], []
 
         image_idx = 1
diff --git a/aidial_adapter_openai/video_generation/azure/types.py b/aidial_adapter_openai/video_generation/azure/types.py
index 26ac0eec..5f65c673 100644
--- a/aidial_adapter_openai/video_generation/azure/types.py
+++ b/aidial_adapter_openai/video_generation/azure/types.py
@@ -1,5 +1,4 @@
 from enum import Enum
-from typing import List
 
 from pydantic import BaseModel, RootModel
 
@@ -41,7 +40,7 @@ class InpaintItem(BaseModel):
 
 
 class InpaintItems(RootModel):
-    root: List[InpaintItem]
+    root: list[InpaintItem]
 
 
 class CreateVideoGenerationRequest(BaseModel):
@@ -68,7 +67,7 @@ def create(
         height: int,
         n_seconds: int | None,
         n_variants: int | None,
-        inpaint_items: List[InpaintItem] | None,
+        inpaint_items: list[InpaintItem] | None,
     ) -> "CreateVideoGenerationRequest":
         inpaint_items_str = (
             InpaintItems(inpaint_items).model_dump_json()
diff --git a/pyproject.toml b/pyproject.toml
index 809675f0..4c3f5e15 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -73,7 +73,7 @@ filterwarnings = [
   "ignore::DeprecationWarning:opentelemetry.instrumentation.dependencies",
   "ignore::DeprecationWarning:pkg_resources",
   # Workaround for this bug: https://github.com/openai/openai-python/issues/1039
-  "ignore:Pydantic serializer warnings:UserWarning:pydantic.*"
+  "ignore:Pydantic serializer warnings:UserWarning:pydantic.*",
 ]
 
 [tool.pyright]
@@ -98,6 +98,7 @@ select = [
   "I",    # flake8-isort
   "S",    # flake8-bandit
   "FURB", # refurb
+  "UP",   #  pyupgrade
 ]
 ignore = [
   "E501", # string literal is too long
diff --git a/tests/integration_tests/base.py b/tests/integration_tests/base.py
index 5030faee..75a7c6da 100644
--- a/tests/integration_tests/base.py
+++ b/tests/integration_tests/base.py
@@ -2,13 +2,10 @@
 
 import functools
 import json
+from collections.abc import Callable, Generator
 from typing import (
     Any,
-    Callable,
-    Dict,
-    Generator,
     Generic,
-    List,
     Literal,
     TypeVar,
     assert_never,
@@ -55,17 +52,17 @@ class Features(ExtraAllowedModel):
 class ModelConfig(ExtraAllowedModel):
     type: Literal["chat", "embedding"]
     overrideName: str | None = None
-    upstreams: List[UpstreamConfig]
+    upstreams: list[UpstreamConfig]
     features: Features = Features()
-    inputAttachmentTypes: List[str] | None = None
+    inputAttachmentTypes: list[str] | None = None
 
 
 class CoreConfig(ExtraAllowedModel):
-    models: Dict[str, ModelConfig]
+    models: dict[str, ModelConfig]
 
     @classmethod
     def from_config(cls, config_path: str):
-        with open(config_path, "r") as f:
+        with open(config_path) as f:
             test_config = json.load(f)
 
         return cls(**test_config)
@@ -102,14 +99,14 @@ class DeploymentConfig(BaseModel, Generic[_T]):
 
     model_name: str
     model_features: Features
-    model_attachments: List[str] | None
+    model_attachments: list[str] | None
 
     upstream_endpoint: str
     upstream_api_key: str | None
     upstream_idx: int | None
 
     @property
-    def upstream_headers(self) -> Dict[str, str]:
+    def upstream_headers(self) -> dict[str, str]:
         headers = {"X-UPSTREAM-ENDPOINT": self.upstream_endpoint}
         if self.upstream_api_key is not None:
             headers["X-UPSTREAM-KEY"] = self.upstream_api_key
@@ -120,7 +117,7 @@ def create_deployments(
         cls,
         core_config: CoreConfig,
         get_deployment_type: Callable[[ModelConfig, str, str], _T],
-    ) -> List[DeploymentConfig[_T]]:
+    ) -> list[DeploymentConfig[_T]]:
         configs = []
         for deployment_id, model_config in core_config.models.items():
             for upstream_index, upstream_config in enumerate(
@@ -187,7 +184,7 @@ def supports_reasoning(self):
 class TestDeployments(BaseModel):
     __test__ = False
 
-    deployments: List[DeploymentConfig[DeploymentType]]
+    deployments: list[DeploymentConfig[DeploymentType]]
     app_config: ApplicationConfig
 
     @classmethod
@@ -224,7 +221,7 @@ def sanitize_id_part(value: Any) -> str:
     """Convert any value to a pytest-safe identifier part."""
     if isinstance(value, bool):
         return "on" if value else "off"
-    if isinstance(value, (int, float)):
+    if isinstance(value, int | float):
         return str(value).replace(".", "p")  # e.g., 0.5 -> 0p5
     if value is None:
         return "none"
diff --git a/tests/integration_tests/chat_completion/response_format.py b/tests/integration_tests/chat_completion/response_format.py
index 7423741b..5f07f2fd 100644
--- a/tests/integration_tests/chat_completion/response_format.py
+++ b/tests/integration_tests/chat_completion/response_format.py
@@ -1,5 +1,5 @@
 import json
-from typing import Any, List
+from typing import Any
 
 from openai.types.chat import ChatCompletionMessageParam
 
@@ -20,7 +20,7 @@ def build_response_format(s: TestSuite) -> None:
     else:
         be_brief = {"max_tokens": 32}
 
-    messages: List[ChatCompletionMessageParam] = [
+    messages: list[ChatCompletionMessageParam] = [
         user("extract name and surname from 'John Doe' in json format")
     ]
 
@@ -31,7 +31,7 @@ def build_response_format(s: TestSuite) -> None:
             response_format={"type": "json_object"},
             **be_brief,
             expected=lambda r: isinstance(
-                _assert_valid_json(r.content), (dict, list)
+                _assert_valid_json(r.content), dict | list
             ),
         )
 
diff --git a/tests/integration_tests/chat_completion/test_case.py b/tests/integration_tests/chat_completion/test_case.py
index e8dcadd6..6d0230dd 100644
--- a/tests/integration_tests/chat_completion/test_case.py
+++ b/tests/integration_tests/chat_completion/test_case.py
@@ -1,7 +1,7 @@
 from __future__ import annotations
 
+from collections.abc import Callable, Iterator
 from dataclasses import dataclass, field
-from typing import Callable, Iterator, List
 
 from openai import NOT_GIVEN, NotGiven
 from openai.types import ReasoningEffort
@@ -27,18 +27,18 @@ class TestCase:
     name: str
     streaming: bool
 
-    messages: List[ChatCompletionMessageParam]
+    messages: list[ChatCompletionMessageParam]
 
     expected: Callable[[ChatCompletionResult], bool] | ExpectedException
 
     max_tokens: int | NotGiven
     max_completion_tokens: int | NotGiven
-    stop: List[str] | NotGiven
+    stop: list[str] | NotGiven
 
     n: int | NotGiven
 
-    functions: List[Function] | NotGiven
-    tools: List[ChatCompletionToolParam] | NotGiven
+    functions: list[Function] | NotGiven
+    tools: list[ChatCompletionToolParam] | NotGiven
     temperature: float | NotGiven
 
     reasoning_effort: ReasoningEffort | NotGiven
@@ -68,19 +68,19 @@ class TestSuite:
 
     deployment_config: DeploymentConfig[ChatCompletionDeploymentType]
     streaming: bool
-    test_cases: List[TestCase] = field(default_factory=list)
+    test_cases: list[TestCase] = field(default_factory=list)
 
     def test_case(
         self,
         *,
         name: str,
-        messages: List[ChatCompletionMessageParam],
+        messages: list[ChatCompletionMessageParam],
         max_tokens: int | NotGiven = NOT_GIVEN,
         max_completion_tokens: int | NotGiven = NOT_GIVEN,
-        stop: List[str] | NotGiven = NOT_GIVEN,
+        stop: list[str] | NotGiven = NOT_GIVEN,
         n: int | NotGiven = NOT_GIVEN,
-        functions: List[Function] | NotGiven = NOT_GIVEN,
-        tools: List[ChatCompletionToolParam] | NotGiven = NOT_GIVEN,
+        functions: list[Function] | NotGiven = NOT_GIVEN,
+        tools: list[ChatCompletionToolParam] | NotGiven = NOT_GIVEN,
         temperature: float | NotGiven = NOT_GIVEN,
         reasoning_effort: ReasoningEffort | NotGiven = NOT_GIVEN,
         response_format: ResponseFormat | NotGiven = NOT_GIVEN,
diff --git a/tests/integration_tests/test_audio.py b/tests/integration_tests/test_audio.py
index 493eaab2..9b6c364d 100644
--- a/tests/integration_tests/test_audio.py
+++ b/tests/integration_tests/test_audio.py
@@ -1,5 +1,5 @@
+from collections.abc import Callable
 from pathlib import Path
-from typing import Callable, List
 from unittest.mock import patch
 
 import openai
@@ -41,7 +41,7 @@ def mock_storage(request):
 
 D = DeploymentConfig[ChatCompletionDeploymentType]
 
-_tts_deployments: List[D] = [
+_tts_deployments: list[D] = [
     d for d in TEST_DEPLOYMENTS_CONFIG.chat_deployments if d.supports_tts
 ]
 
@@ -58,7 +58,7 @@ def tts_deployment(request) -> D:
         pytest.skip("No TTS deployments were found")
 
 
-_stt_deployments: List[D] = [
+_stt_deployments: list[D] = [
     d for d in TEST_DEPLOYMENTS_CONFIG.chat_deployments if d.supports_stt
 ]
 
diff --git a/tests/integration_tests/test_chat_completion.py b/tests/integration_tests/test_chat_completion.py
index 77e0ddc9..a0d0aaed 100644
--- a/tests/integration_tests/test_chat_completion.py
+++ b/tests/integration_tests/test_chat_completion.py
@@ -1,6 +1,6 @@
 import logging
 import re
-from typing import Generator, List
+from collections.abc import Generator
 
 import openai
 import pytest
@@ -34,7 +34,7 @@
 
 
 def create_test_cases(
-    builders: List[TestSuiteBuilder],
+    builders: list[TestSuiteBuilder],
 ) -> Generator[TestCase, None, None]:
     for streaming in (False, True):
         for deployment in TEST_DEPLOYMENTS_CONFIG.chat_deployments:
diff --git a/tests/integration_tests/test_embeddings.py b/tests/integration_tests/test_embeddings.py
index 539c4d57..3014a76e 100644
--- a/tests/integration_tests/test_embeddings.py
+++ b/tests/integration_tests/test_embeddings.py
@@ -1,5 +1,3 @@
-from typing import List
-
 import openai
 import pytest
 
@@ -15,7 +13,7 @@
 
 D = DeploymentConfig[EmbeddingsDeploymentType]
 
-_deployments: List[D] = list(TEST_DEPLOYMENTS_CONFIG.embedding_deployments)
+_deployments: list[D] = list(TEST_DEPLOYMENTS_CONFIG.embedding_deployments)
 
 
 @pytest.fixture
diff --git a/tests/integration_tests/test_image_generation.py b/tests/integration_tests/test_image_generation.py
index 53e085b2..8648f61a 100644
--- a/tests/integration_tests/test_image_generation.py
+++ b/tests/integration_tests/test_image_generation.py
@@ -1,5 +1,5 @@
+from collections.abc import Callable
 from pathlib import Path
-from typing import Callable, List
 from unittest.mock import patch
 
 import openai
@@ -33,7 +33,7 @@ def mock_storage(request):
 
 D = DeploymentConfig[ChatCompletionDeploymentType]
 
-_deployments: List[D] = [
+_deployments: list[D] = [
     d
     for d in TEST_DEPLOYMENTS_CONFIG.chat_deployments
     if d.model_features.imageGenerationSupported
@@ -54,7 +54,7 @@ def imagen_deployment(request) -> D:
 
 @pytest.fixture
 def vision_deployment() -> D:
-    vision_deployments: List[D] = [
+    vision_deployments: list[D] = [
         d
         for d in TEST_DEPLOYMENTS_CONFIG.chat_deployments
         if d.supports_vision and not d.supports_reasoning
diff --git a/tests/integration_tests/test_video_generation.py b/tests/integration_tests/test_video_generation.py
index f8dda8b4..a4a46951 100644
--- a/tests/integration_tests/test_video_generation.py
+++ b/tests/integration_tests/test_video_generation.py
@@ -1,5 +1,5 @@
+from collections.abc import Callable
 from pathlib import Path
-from typing import Callable, List
 from unittest.mock import patch
 
 import openai
@@ -33,7 +33,7 @@ def mock_storage(request):
 
 D = DeploymentConfig[ChatCompletionDeploymentType]
 
-_deployments: List[D] = [
+_deployments: list[D] = [
     d
     for d in TEST_DEPLOYMENTS_CONFIG.chat_deployments
     if d.supports_video_generation
diff --git a/tests/unit_tests/test_caching.py b/tests/unit_tests/test_caching.py
index dc91bb47..8abcffef 100644
--- a/tests/unit_tests/test_caching.py
+++ b/tests/unit_tests/test_caching.py
@@ -1,6 +1,5 @@
 import dataclasses
 import json
-from typing import List
 
 import httpx
 import pytest
@@ -14,7 +13,7 @@
 from tests.utils.stream import OpenAIStream, single_choice_chunk
 
 
-def _mock_response(upstream_url: str, stream: bool, chunks: List[dict]):
+def _mock_response(upstream_url: str, stream: bool, chunks: list[dict]):
     mock_stream = OpenAIStream(*chunks)
     if stream:
         respx.post(upstream_url).respond(
diff --git a/tests/unit_tests/test_discard_messages.py b/tests/unit_tests/test_discard_messages.py
index 35ee8b55..08058644 100644
--- a/tests/unit_tests/test_discard_messages.py
+++ b/tests/unit_tests/test_discard_messages.py
@@ -1,5 +1,3 @@
-from typing import List, Tuple
-
 import pytest
 from aidial_sdk.exceptions import HTTPException as DialException
 
@@ -13,21 +11,21 @@
     TruncatedTokens,
 )
 
-PlainTextMessages = List[dict]
+PlainTextMessages = list[dict]
 MaxPromptTokens = int
-TestCase = Tuple[
+TestCase = tuple[
     PlainTextMessages,
     MaxPromptTokens,
-    Tuple[PlainTextMessages, DiscardedMessages],
+    tuple[PlainTextMessages, DiscardedMessages],
 ]
 
 
 async def plain_text_truncate_prompt(
     request: dict,
-    messages: List[dict],
+    messages: list[dict],
     max_prompt_tokens: int,
     tokenizer: Tokenizer,
-) -> Tuple[List[dict], DiscardedMessages, TruncatedTokens]:
+) -> tuple[list[dict], DiscardedMessages, TruncatedTokens]:
     (msgs, disc, tokens) = await multi_modal_truncate_prompt(
         request=request,
         messages=[MultiModalMessage(raw_message=m) for m in messages],
@@ -39,7 +37,7 @@ async def plain_text_truncate_prompt(
     return (msgs, disc, tokens)
 
 
-normal_cases: List[TestCase] = [
+normal_cases: list[TestCase] = [
     (
         [],
         3,
@@ -124,8 +122,8 @@ async def plain_text_truncate_prompt(
 ]
 
 ErrorMessage = str
-error_cases: List[
-    Tuple[
+error_cases: list[
+    tuple[
         PlainTextMessages,
         MaxPromptTokens,
         ErrorMessage,
@@ -159,9 +157,9 @@ async def plain_text_truncate_prompt(
 
 @pytest.mark.parametrize("messages, max_prompt_tokens, response", normal_cases)
 async def test_discarded_messages_without_error(
-    messages: List[dict],
+    messages: list[dict],
     max_prompt_tokens: int,
-    response: Tuple[List[dict], List[int]],
+    response: tuple[list[dict], list[int]],
 ):
     tokenizer = Tokenizer(model="gpt-4")
     (
@@ -178,7 +176,7 @@ async def test_discarded_messages_without_error(
     "messages, max_prompt_tokens, error_message", error_cases
 )
 async def test_discarded_messages_with_error(
-    messages: List[dict],
+    messages: list[dict],
     max_prompt_tokens: int,
     error_message: str,
 ):
diff --git a/tests/unit_tests/test_errors.py b/tests/unit_tests/test_errors.py
index 71d65390..fcd71077 100644
--- a/tests/unit_tests/test_errors.py
+++ b/tests/unit_tests/test_errors.py
@@ -1,5 +1,6 @@
 import json
-from typing import Any, AsyncIterable, AsyncIterator, Callable
+from collections.abc import AsyncIterable, AsyncIterator, Callable
+from typing import Any
 from unittest.mock import patch
 
 import httpx
diff --git a/tests/unit_tests/test_gpt_audio.py b/tests/unit_tests/test_gpt_audio.py
index 33c0b6ba..b3a35e5a 100644
--- a/tests/unit_tests/test_gpt_audio.py
+++ b/tests/unit_tests/test_gpt_audio.py
@@ -1,6 +1,7 @@
 """Unit tests for GPT-4o audio completions support."""
 
-from typing import Any, AsyncIterator, Callable
+from collections.abc import AsyncIterator, Callable
+from typing import Any
 
 import pytest
 
diff --git a/tests/unit_tests/test_image_tokenization.py b/tests/unit_tests/test_image_tokenization.py
index 05d3117b..d351c587 100644
--- a/tests/unit_tests/test_image_tokenization.py
+++ b/tests/unit_tests/test_image_tokenization.py
@@ -1,11 +1,9 @@
-from typing import List, Tuple
-
 import pytest
 
 from aidial_adapter_openai.utils.image_tokenizer import GPT4O_IMAGE_TOKENIZER
 from aidial_adapter_openai.utils.resource.image import ImageDetail
 
-test_cases: List[Tuple[int, int, ImageDetail, int]] = [
+test_cases: list[tuple[int, int, ImageDetail, int]] = [
     (1, 1, "auto", 85),
     (1, 1, "high", 170 * 1 + 85),
     (100, 100, "low", 85),
diff --git a/tests/utils/dictionary.py b/tests/utils/dictionary.py
index e5f7b87a..1f964955 100644
--- a/tests/utils/dictionary.py
+++ b/tests/utils/dictionary.py
@@ -1,4 +1,4 @@
-from typing import Iterable
+from collections.abc import Iterable
 
 
 def exclude_keys(d: dict, keys: Iterable[str]) -> dict:
diff --git a/tests/utils/json.py b/tests/utils/json.py
index 896a2af1..7299682a 100644
--- a/tests/utils/json.py
+++ b/tests/utils/json.py
@@ -6,7 +6,7 @@ def match_objects(expected: Any, actual: Any) -> bool:
         assert sorted(expected.keys()) == sorted(actual.keys())
         for k, v in expected.items():
             match_objects(v, actual[k])
-    elif isinstance(expected, (tuple, list)):
+    elif isinstance(expected, tuple | list):
         assert len(expected) == len(actual)
         for i in range(len(expected)):
             match_objects(expected[i], actual[i])
diff --git a/tests/utils/openai.py b/tests/utils/openai.py
index 3a76603b..7a8d9eb6 100644
--- a/tests/utils/openai.py
+++ b/tests/utils/openai.py
@@ -1,5 +1,6 @@
 import json
-from typing import Any, Callable, List
+from collections.abc import Callable
+from typing import Any
 
 from aidial_sdk.utils.merge_chunks import (
     cleanup_indices,
@@ -53,13 +54,13 @@ def ai_function(
 
 
 def ai_tools(
-    tool_calls: List[ChatCompletionMessageToolCallParam],
+    tool_calls: list[ChatCompletionMessageToolCallParam],
 ) -> ChatCompletionAssistantMessageParam:
     return {"role": "assistant", "tool_calls": tool_calls}
 
 
 def user(
-    content: str | List[ChatCompletionContentPartParam],
+    content: str | list[ChatCompletionContentPartParam],
     **kwargs,
 ) -> ChatCompletionUserMessageParam:
     return {"role": "user", "content": content, **kwargs}  # type: ignore
@@ -186,7 +187,7 @@ def content(self) -> str:
         return self.message.content or ""
 
     @property
-    def contents(self) -> List[str]:
+    def contents(self) -> list[str]:
         return [
             choice.message.content or "" for choice in self.response.choices
         ]
@@ -200,10 +201,10 @@ def function_call(self) -> FunctionCall | None:
         return self.message.function_call
 
     @property
-    def tool_calls(self) -> List[ChatCompletionMessageToolCall] | None:
+    def tool_calls(self) -> list[ChatCompletionMessageToolCall] | None:
         return self.message.tool_calls
 
-    def content_contains_all(self, matches: List[Any]) -> bool:
+    def content_contains_all(self, matches: list[Any]) -> bool:
         return all(
             str(match).lower() in self.content.lower() for match in matches
         )
@@ -213,14 +214,14 @@ async def chat_completion(
     client: AsyncAzureOpenAI,
     *,
     deployment_id: str,
-    messages: List[ChatCompletionMessageParam],
+    messages: list[ChatCompletionMessageParam],
     stream: bool,
-    stop: List[str] | NotGiven = NOT_GIVEN,
+    stop: list[str] | NotGiven = NOT_GIVEN,
     max_completion_tokens: int | NotGiven = NOT_GIVEN,
     max_tokens: int | NotGiven = NOT_GIVEN,
     n: int | NotGiven = NOT_GIVEN,
-    functions: List[Function] | NotGiven = NOT_GIVEN,
-    tools: List[ChatCompletionToolParam] | NotGiven = NOT_GIVEN,
+    functions: list[Function] | NotGiven = NOT_GIVEN,
+    tools: list[ChatCompletionToolParam] | NotGiven = NOT_GIVEN,
     temperature: float | NotGiven = NOT_GIVEN,
     reasoning_effort: ReasoningEffort | NotGiven = NOT_GIVEN,
     response_format: ResponseFormat | NotGiven = NOT_GIVEN,
@@ -246,7 +247,7 @@ async def get_response() -> ChatCompletion:
         )
 
         if isinstance(response, AsyncStream):
-            chunks: List[dict] = [
+            chunks: list[dict] = [
                 {}
             ]  # workaround for https://github.com/epam/ai-dial-sdk/pull/269
             async for chunk in response:
@@ -300,7 +301,7 @@ def is_valid_function_call(
 
 
 def is_valid_tool_call(
-    calls: List[ChatCompletionMessageToolCall] | None,
+    calls: list[ChatCompletionMessageToolCall] | None,
     tool_call_idx: int,
     check_tool_id: Callable[[str], bool],
     expected_name: str,
diff --git a/tests/utils/storage.py b/tests/utils/storage.py
index 0b852929..01c3d331 100644
--- a/tests/utils/storage.py
+++ b/tests/utils/storage.py
@@ -1,7 +1,6 @@
 import mimetypes
 import os
 from pathlib import Path
-from typing import List
 from urllib.parse import urlparse
 
 from pydantic import SecretStr
@@ -42,7 +41,7 @@ async def download_file(self, link: str) -> bytes:
 
 class MockFileStorage(FileStorage):
     root_dir: Path
-    files: List[Path]
+    files: list[Path]
 
     @classmethod
     def create(cls, root_dir: Path) -> "MockFileStorage":
diff --git a/tests/utils/stream.py b/tests/utils/stream.py
index d3db7e33..82c1eb63 100644
--- a/tests/utils/stream.py
+++ b/tests/utils/stream.py
@@ -1,5 +1,6 @@
 import json
-from typing import Any, Callable, List
+from collections.abc import Callable
+from typing import Any
 
 import httpx
 
@@ -9,7 +10,7 @@
 
 
 class OpenAIStream:
-    chunks: List[dict]
+    chunks: list[dict]
 
     def __init__(self, *chunks: dict):
         self.chunks = list(chunks)
@@ -55,7 +56,7 @@ def chunk(
     object_: str = "chat.completion.chunk",
     created: int = 1695940483,
     model: str = "gpt-4",
-    choices: List[dict],
+    choices: list[dict],
     usage: dict | None = None,
     **kwargs,
 ) -> dict:
@@ -125,7 +126,7 @@ def many_choices_chunk(
     created: int = 1695940483,
     model: str = "gpt-4",
     usage: dict | None = None,
-    choices: List[dict] | None = None,
+    choices: list[dict] | None = None,
     **kwargs,
 ) -> dict:
     return chunk(