Keep getting rate limited on Zen

[thefrana]

Description

I am using OpenCode Zen with Kimi-K2.5 but I am getting rate limited even after a few queries. I am not even using the free version but it seems paying for the model does not help with anything. I am reaching up to 8th or more attempt. Any ideas what I should do?

Plugins

None

OpenCode version

1.1.59

Steps to reproduce

1. Open OpenCode
2. Start using Zen
3. Rate limits blocks you

Screenshot and/or share link

No response

Operating System

macOS

Terminal

Ghostty

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• API Rate Limit Not Properly Handled - Model Stops Responding After Few Queries #11104: API Rate Limit Not Properly Handled - Model Stops Responding After Few Queries
• Big Pickle model loops with "too many requests" error in high mode (free trial) #10404: Big Pickle model loops with "too many requests" error in high mode
• 提示： Rate limit exceeded ，please try again 是什么情况的现在 #11242: Rate limit exceeded error messages with Zen models
• Implement exponential back-off when hitting rate limits #1712: Implement exponential back-off when hitting rate limits (related feature request)

These issues describe similar rate limiting problems with OpenCode Zen and models. Reviewing these might help you find existing workarounds or follow-up on the core issue.

[vitobotta]

I am also having this super frustrating issue. I am using paid Kimi K2.5, yet it happens quite often even when working on a single task at a time.

Is there any support at all for this service?

[thefrana]

I think the problem is Fireworks (IFAIK used in Zen). The metrics on OpenRouter says they are one of the worst providers. Reliability was 50% for Kimi K2.5. I hope Zen will eventually switch to something more stable because this is unbearable.

[vitobotta]

I switched to OpenRouter, selected the top 5 performing providers and it's working great. I can work on multiple tasks at the same time without any issues and the speed is basically the same, or at least I can't see any difference so far.

[thefrana]

Thanks. I also tried OpenRouter and it highly varies which provider it selects.

BTW, I found it very interesting that Fireworks does not even bother to show Kimi K2.5 on their status page: https://status.fireworks.ai/

Apparently, it looks like they resign on updating the status website with newer models or the models have such bad uptime they do not even want to show it.

[vitobotta]

I recommend you only enable the best performing providers in the settings, instead of letting OpenRouter pick any provider available, since some of them are slow.