@Seraphina thats not unknown, the toast spells it out, your AI Agent node is pinned at an old version (2.2) that doesnt support the model youve wired in. n8n locks a nodes version when its added, so swap it for a fresh one, delete the AI Agent and drop in a new AI Agent node, reconnect your openai model, memory and tools and the main in/out. the new one instantiates at the current version and the model works. if the new node still comes in at 2.2, then your n8n instance itself is behind and youd update n8n to get the newer agent.
Can anyone tell me why my OpenAI Chat Model is failing? It was working perfectly fine before. I also checked each node, it’s updated and supports 2.0. OpenRouter Chat Model is working, Gemini is working but why Open AI is always failing.
Did anyone face the same issue?
The best solution: Select AI Agent Node. In the Node panel on the top right, use the version selector and update to the latest version of that Node. The Node will keep all current connections. The OpenAI Chat Model is expected to work after the update.
This is caused by the type version of the Agent Node being updated with n8n, while the existing workflows are still on the previous version until you update manually.
@Seraphina since openrouter and gemini both work and only openai fails, that points at the openai side specifically not your agent setup, usually the api key or the account being out of quota/credits (that lines up with it working before then stopping). whats the actual error on the openai chat model node when it runs? open it, execute, and paste what the red error says, that tells us if its a 401 key issue, a 429 quota/billing one, or a model-not-found.
If the error is exactly what’s being shown on the screenshot, i would not have posted here.
@Seraphina openrouter and gemini up but only openai dead is almost always openai out of credits, check billing on platform.openai.com. if theres balance there, paste the exact red error off the openai node, cant tell key vs model otherwise.
Fair enough. If you’ve rebuilt the node and it continues to throw the same error, it doesn’t appear to be the obvious version bump.
Here are some less obvious things that cause this even on a fresh node:
• Stale type Version in the JSON file. Sometimes the canvas updates the node, but the workflow JSON updates do not. You should export the workflow and check the typeVersion for the AI Agent node directly in the JSON. If it is not the latest version, you can update it in the JSON file and then reimport the workflow.
• One step down the Agent node is the OpenAI sub-node. Even though the error references the Agent node, the incompatibility could be with the OpenAI Chat Model sub-node and the version that the Agent node is incompatible with. You should delete the OpenAI sub-node and re-add it (do not do this for the Agent node).
• To be clear, this is a versioning issue with your self-hosting the n8n instance. If you are self-hosting and are a few versions behind, the latest Agent typeVersion may not even be available in your instance. What version of n8n are you using?