H Lohaus

Yes instead of Cookies, you can pass the access_token. Where did you get the token from?

You have still errors? ``` response = g4f.ChatCompletion.create( model=g4f.models.gpt_35_turbo, provider=g4f.Provider.OpenaiChat, messages=[{"role": "user", "content": "Create me a image with a lion."}], access_token="eyJhbGciOiJ.........", auth=True ) ``` Works for me

i fixed the issue with conversationSignature and Failed to create conversation. if i clear the cookie jar, the second request works.

You can now pass proxy at create: ...create(...., proxy="http://host:post")

create_async doesn't support stream=True,

I have included the AsyncClient with examples in the documentation.

Try out GPT 3.5 or "default": model=g4f.models.gpt_35_turbo, or model=g4f.models.default, Free GPT 4 providers are rare

You can pass argument `proxy="socks://localhost:1337" to many, but not all providers.

Did you test other providers? The providers from the default list are almost not working. And you should try use a other IP address. I used yesterday a VPN for...

Curl-cffi doesn't support Windows 32 bit