Im Currently Running into an issue where my Sub Agent Tool is failing to run because “It doesnt support the Model” (See screenshots). Ive already updated my instance to the latest version and in the Agent node Settings it says “version (latest)”. Ive tried to run GPT 5 with a “high” thinking settings but have also tried GPT40 latest. However I keep running into that issue - What is the Workaround? How do I Update my Agent Node if it states its the latest version?
Thanks on Advance for the Help!
Hey @Jonas_Hermann Looks like this is most likely a node version mismatch inside your workflow, not an issue with your n8n instance or the model itself.
Even if n8n says everything is “latest,” older Agent / Sub-Agent Tool nodes don’t automatically upgrade. When that happens, they can’t run newer models and throw the “not supported in 2.2” error you’re seeing.
The fix is pretty simple:
1. Delete your existing AI Agent + Sub-Agent Tool nodes (the ones already in the workflow).
2. Add fresh ones from the node picker.
3. Reconnect your chat model (gpt-4o or similar tends to work best with tools).
This pulls in the newest internal node version, and in most cases the error disappears right away.
If you still hit the same message even after recreating the nodes, then it might be a new bug in that case, sharing a small workflow example and your n8n version would help narrow it down.
Hope that helps! Let me know if you want me to check anything else in your setup.
So I’ve deleted both my Sub and my Main Agent Node and replaced it with a “fresh” one, but the problem still remains the same 
Since my Main Agent is actually a “Message an Assistant” Node from OpenAI, I’ve also replaced this with a native AI Agent Node but still it shows the same Error - Although the Main Agent Node runs perfectly with the same Model (Both connected to the same Model Node)
I’ve been trying to even run super old models, and it still doesn’t work.
Here is a screenshot of the Current Workflow - Indicating the Error in the Subagent
Hey @Jonas_Hermann sorry to hear that the issue still persists..
lets rewire, what’s happening is that your sub-agent is still being fed a model the Agent Tool node can’t work with, even though the “main” agent runs without issues. The node labeled “CRO Atlas Reasoning” looks like a reasoning-type model (or a custom wrapper of one), and those models don’t support tool-calls, which the Sub-Agent Tool depends on. That’s why the error specifically appears in the sub-agent and not in the main one.
Agents and Sub-Agent Tools only work with Chat Model-type models such as:
gpt-4o
gpt-4.1
Other vendor chat models (Anthropic, Groq, etc.)
Reasoning models, or anything that uses the new /v1/responses format, can’t be used inside Agent nodes. If you need them, they have to be triggered via an HTTP Request node instead.
Swap the “CRO Atlas Reasoning” node in the sub-agent chain for a normal OpenAI Chat Model → gpt-4o, then run it again.
If it works with gpt-4o and fails with your reasoning model, you’ve found the root cause.
Then we might be dealing with a small regression in the current Agent/Tool logic. In that case, send a tiny workflow export (trigger → agent → model → sub-agent tool) plus your n8n version, and I can try to reproduce it on my side.
Let me know what happens after switching the model, that will narrow it down immediately.
