Open AI Rate limits

[terse crater]

The "exceeded OpenAI rate limits" error on your n8n workflow using free models on OpenRouter likely stems from upstream providers hitting request limits or temporary overload. OpenRouter routes requests to various providers, and when one is overloaded, it may return 429 errors ("Too many requests"). This can happen even with credits and verified credentials. OpenRouter can recover by routing to other healthy providers if your request filters allow it. You can learn more about provider routing and load balancing here: https://openrouter.ai/docs/provider-routing

If changing the fallback model causes errors, ensure your workflow supports the new model’s parameters and availability.

For persistent or random issues, reaching out to your Discord server admin or OpenRouter support is recommended, as it may be related to provider status or maintenance.

Are you using any specific provider filters or custom routing in your setup?

-# This is an automated response. Use !mute to stop the bot from responding.

[violet elk]

I understand but now when I change model to Gemma 4b:free I get another error saying that it can't find the resource.

It keeps going back and forth between OpenAi rate limit

There has to be a way to fix this behavior, isn't there?

[terse crater]

The "resource not found" error with the Gemma 4B free model likely means the model name or endpoint is incorrect or temporarily unavailable. OpenRouter sometimes faces upstream rate limits (429 errors) on free models like Gemma 4B due to provider overload.

To fix this:

• Double-check the exact model name and version in your API calls. For Gemma 4B free, use the correct identifier from OpenRouter’s model list, e.g., "google/gemma-3-4b-it:free" or the exact name shown on OpenRouter.
• Avoid hardcoding model names; use the updated list from OpenRouter’s dashboard.
• If you apply provider filters or strict routing, loosen them to allow fallback to healthy providers.
• Consider spreading requests over time to reduce hitting burst rate limits.
• For persistent rate limit errors, it’s best to contact your Discord server admin or OpenRouter support for the latest status.

You can review load balancing and routing details here: https://openrouter.ai/docs/provider-routing

Which platform is your workflow running on? I can help with tailored setup advice.

-# This is an automated response. Use !mute to stop the bot from responding.

[violet elk]

How can I reach the admin?

[fringe mango]

!mute

[terse crater]

🔇 Thread muted. The bot will no longer respond here.

[fringe mango]

what model are you using?

[violet elk]

I am currently using qwen3-4b:free and previously tested with deepseek/deepseek-chat-v3-0324:free

Both models seemed to be working but now it appears that open router may have been falling back on an early version of OpenAI

[sullen briar]

so, if this is Gemma4b:free, the error is right

it doesn't support tool calling.

if it did, you'd see it have a tag under supported parameters that says Tools

@violet elk what it looks like on o3-mini... shows Tools tag

try with openrouter/horizon-alpha

it is free

[violet elk]

Ok, let me check. I am using 'qwen/qwen3-4b:free' that does support tools from what I am seeing

'openrouter/horizon-alpha' is working

thank you for the advice, I appreciate it.

[wanton hawk]

[ERROR: OpenAl: Rate limit reached] this error message text is not actually from OpenRouter, looks like n8n uses that instead of showing the actual message, which is confusing because OpenAI isn't actually involved.

openrouter/horizon-alpha is temporary, completely unlimited free stealth model and will disappear soon. regular free models do have some sort of undocumented(?) RPM limit, which you're liking hitting with an agentic workflow.

free models aren't intended for production use. you'll have to use paid models after horizon is gone.