We noticed that usage dropped significantly after Grok became the only available option, so we launched a new channel — opencode zen.
We have implemented support on a small scale for using third-party platforms such as DeepSeek Harness to call opencode models.
To address opencode’s IP-based rate limits, we are building a proxy pool, which we expect to launch in the coming days.
Are you taking the route to my station?
Take a look at this model’s performance—you’ll probably give up…
What is the prompt?
Create an HTML page containing a 2D animation of a pelican riding a bicycle, drawn with SVG. You don’t need any tests.
How do I use your website?!
Obviously, his Grok public API site is https://api.lezi-ai.bond
If you want to use the Opencode Zen models right now, you can use my https://api.tokentoken.bond/
The reason usage of this 乐子AI site dropped so noticeably after it was reduced to only Grok isn’t that Grok is bad.
It may simply be that the model is too unstable. Your account pool is too small, so it keeps throwing errors, and the time to the first token is too long. Most of the time, you’re just waiting for a request to complete…
Mimo can draw on my end.
Got it.
opencode zen keeps returning 429. I don’t know what’s going on.
I’ve tried switching to five different IPs, but it’s still the same.
Is yours an airport IP?
A mixed pool (residential broadband, commercial, educational, and IDC, mostly commercial)
There was almost none of this in the morning
But now there is
Probably upstream of opencode.
