Thanks to Microsoft, the generous benefactor, for letting me use GPT 5.6 so smoothly 😭🙏

rt, I’ve really enjoyed using it these past few days (just on one account)

I noticed that during the day there are basically no restrictions—it’s only at night that I tend to get hit with 429s. (Maybe because Americans are sleeping during the day?)

​Total token usage: 31.16M = 31.16 million tokens

​Input tokens (not cached): 15,542,360 ≈ 15.54 million tokens

​Cache hits (Cached): 15.53M = 15.53 million tokens

​Output tokens: 88,710 ≈ 88.7 thousand tokens

Official API total original price (counting only GPT 5.6): about $43.62 (roughly 310 RMB) :ac01:

Special thanks again to the American company, Microsoft! :xhj16:

45 Likes

Can it be deployed on a server? I want to set it up on a server as a translation API.

2 Likes

It’s doable, but watch the IP. Just don’t get Huge-hard pissed off.

1 Like

man tool calls work??

2 Likes

Of course you can!

2 Likes

Take a look

1 Like

Feels pretty good.

1 Like

Is Linux also okay? Then I’ll deploy one there. As for the IP, would it work if I set up a dynamic proxy?

2 Likes

That’s fine, but I recommend not making the frontend accessible from the public internet—I can’t guarantee the project is free of vulnerabilities.

2 Likes

Kayan, I love you!!!
Mwah!

4 Likes

But the failover isn’t fully set up yet…:xhj17:

2 Likes

Which cli or ide u use?? I used opencode but it doesn’t work for tool calls or file modifications

3 Likes

Huh? But I’m using it in opencode. It works fine…

1 Like

Claude Code works just as well…

I will try again and let u know

I just used Codex directly. It can be used, but it doesn’t work.

I just feel like there are still some problems I don’t know how to solve. Every time I call this API to generate a document, the document can’t be opened, and even after logging in, it still shows an error.

Sometimes this issue occurs: “I cannot confirm completion because no matching tool results were returned. No external action has been verified.”

1 Like

Hey, can I ask you something? After deploying this, when I switch the ChatGPT app to this custom model and ask a question, why does it repeat the answer several times?

Wow, that’s amazing! How did you use it? I don’t think I saw it when I was looking through older posts. :xhj17: