diff --git a/pyproject.toml b/pyproject.toml
index 8221957..2daf7f8 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -12,16 +12,16 @@ dependencies = [
     "promplate-trace[langfuse,langsmith]==0.3.0dev5",
     "python-box~=7.2.0",
     "pydantic-settings~=2.6.0",
-    "httpx[http2]~=0.27.0",
+    "httpx[http2]~=0.28.0",
     "promptools[stream,validation]~=0.1.3.3",
-    "fake-useragent~=1.5.1",
+    "fake-useragent~=2.0.0",
     "html2text~=2024.2.26",
     "beautifulsoup4~=4.12.3",
     "rich~=13.9.0",
-    "zhipuai~=2.1.0.20240521",
-    "anthropic~=0.37.1",
+    "zhipuai~=2.1.5.20241203",
+    "anthropic~=0.40.0",
     "dashscope~=1.20.2",
-    "logfire[fastapi,system-metrics,httpx]~=2.1.1",
+    "logfire[fastapi,system-metrics,httpx]~=2.6.0",
 ]
 
 [tool.pdm]
diff --git a/src/logic/tools/fetch.py b/src/logic/tools/fetch.py
index 57c38f6..8b216f9 100644
--- a/src/logic/tools/fetch.py
+++ b/src/logic/tools/fetch.py
@@ -4,13 +4,13 @@
 from typing import cast
 
 from bs4 import BeautifulSoup
-from fake_useragent import UserAgent
+from fake_useragent.fake import FakeUserAgent
 from html2text import html2text
 from httpx import AsyncClient, HTTPError
 
 from .base import AbstractTool
 
-ua = UserAgent(min_percentage=0.5)
+ua = FakeUserAgent(min_percentage=0.5)
 
 
 class Browser(AbstractTool):
diff --git a/src/routes/openai.py b/src/routes/openai.py
index 2830af7..6880dea 100644
--- a/src/routes/openai.py
+++ b/src/routes/openai.py
@@ -33,7 +33,7 @@ class ChatInput(ChainInput):
 
     @property
     def config(self):
-        return self.model_dump(exclude_unset=True)
+        return self.model_dump(exclude_unset=True, exclude={"messages"})
 
 
 def mix_config(r: Request, data: ChatInput):
diff --git a/src/utils/llm/__init__.py b/src/utils/llm/__init__.py
index ae7a46a..635eba0 100644
--- a/src/utils/llm/__init__.py
+++ b/src/utils/llm/__init__.py
@@ -68,11 +68,14 @@
     "llama3-70b-8192",
     "llama-3.1-8b-instant",
     "llama-3.1-70b-versatile",
+    "llama-3.1-70b-specdec",
     "llama-3.1-405b-reasoning",
-    "llama-3.2-1b-text-preview",
-    "llama-3.2-3b-text-preview",
-    "llama-3.2-11b-text-preview",
-    "llama-3.2-90b-text-preview",
+    "llama-3.2-1b-preview",
+    "llama-3.2-3b-preview",
+    "llama-3.2-11b-vision-preview",
+    "llama-3.2-90b-vision-preview",
+    "llama-3.3-70b-versatile",
+    "llama-3.3-70b-specdec",
     "llama3.1-8b",
     "llama3.1-70b",
     "mixtral-8x7b-32768",
diff --git a/src/utils/llm/groq.py b/src/utils/llm/groq.py
index c6509ae..41a956e 100644
--- a/src/utils/llm/groq.py
+++ b/src/utils/llm/groq.py
@@ -16,6 +16,7 @@
 @link_llm("llama3-")
 @link_llm("llama-3.1")
 @link_llm("llama-3.2")
+@link_llm("llama-3.3")
 @link_llm("mixtral")
 class Groq(AsyncChatOpenAI):
     async def complete(self, prompt: str | list[Message], /, **config):
diff --git a/src/utils/response.py b/src/utils/response.py
index 3ddda44..f1dc309 100644
--- a/src/utils/response.py
+++ b/src/utils/response.py
@@ -1,4 +1,3 @@
-from operator import call
 from traceback import print_exc
 
 from fastapi.responses import PlainTextResponse, StreamingResponse
@@ -14,10 +13,9 @@ async def make_response(stream: AsyncContentStream, media_type="text/plain"):
         print_exc()
         return PlainTextResponse("\n".join(map(str, e.args)).strip(), getattr(e, "status_code", 500))
 
-    @call
     async def _():
         yield first_chunk
         async for i in it:
             yield i
 
-    return StreamingResponse(_, media_type=media_type)
+    return StreamingResponse(_(), media_type=media_type)