Model overloaded but I changed the model

[FierdCS]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
• [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
• Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

Yes, I recently had this workflow I called via an API key. However, the Gemini LLM I was using overloaded and so I switched the LLM on the cloud UI to another version of the model. I then proceeded to try running my script connected to the workflow with the valid API key but I still got the model overloaded error even though the workflow works on the cloud UI. Here is the error I'm getting with the model:

{'task_id': 'hiding for privacy', 'workflow_run_id': 'hiding for privacy', 'data': {'id': 'hiding for privacy', 'workflow_id': 'hiding for privacy', 'status': 'failed', 'outputs': {}, 'error': 'req_id: 24f2fb6e70 PluginInvokeError: {"args":{},"error_type":"ServerError","message":"503 UNAVAILABLE. {'error': {'code': 503, 'message': 'The model is overloaded. Please try again later.', 'status': 'UNAVAILABLE'}}"}', 'elapsed_time': 1.882283, 'total_tokens': 0, 'total_steps': 2, 'created_at': 1761581561, 'finished_at': 1761581563}}

2. Additional context or comments

No response

[dosubot[bot]]

The 503 "model overloaded" error is surfaced directly from the LLM provider or plugin and can persist in API-triggered workflows even after switching models in the UI if the workflow nodes don't have proper error handling or retry strategies configured. Dify's error handling system lets you set per-node strategies (like retry, fallback, or fail branch) to make workflows more resilient to transient failures like model overloads—these settings are especially important after recent architectural changes in v2.x. If your workflow works in the UI but not via API, it's possible the API-triggered run is still referencing outdated configuration, or the change hasn't fully propagated due to a temporary delay or caching at the provider/plugin level. There are also known issues with Gemini and other plugin-based models that can cause discrepancies between UI and API behavior, so double-check your model and credential configuration, and consider adding error handling to your workflow nodes for better resilience. More details and troubleshooting steps are available in the error handling guide and related issue discussions.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}