What it is
Your AI chat agents (the Conversation AI agents in TraxionLab) each run on an AI model you pick from a menu. The release says GPT 6 Sol answered more accurately than GPT 4.1 in their own testing, at about the same cost per reply. Switch now and test, even if your agent seems fine. A model change can make your agent sound a little different, and you want to catch that in a test chat, not in a real conversation with a lead.
GPT 6 Sol is recommended
GPT 6 Sol is now first in the Model menu and tagged Recommended.
Existing agents stay put
Existing agents keep the model you picked. Nothing switches on its own, except agents that were on a model that has already been retired.
GPT-5 already moved
GPT-5 and GPT-5 Mini are retired. Agents that used them already run on GPT 6 Sol, so there's nothing to do there.
GPT 4.1 is going away
OpenAI, the company that makes it, is shutting GPT 4.1 down, and it will be removed from the menu. As of Nov 10 it's still in the menu, with no retirement notice on the screen, and no removal date has been announced yet.
Who it's for
Any account with an AI agent replying to customers, especially ones set up more than a few months ago, since those are the most likely to still be on GPT 4.1. It applies to prompt-based chat agents. It doesn't say anything about voice agents.
- Text, chat, and DM agents
- Agents set up months ago
- Prompt-based chat agents
Set it up in TraxionLab
You switch each agent yourself. Anything marked Check in your account is a spot where your screen may look a little different. If you'd rather we do it for you, reply to our email.
Part 1 · Find which model each agent uses
Log in at app.traxionlab.com and go to AI Agents → Conversation AI.
Open the Agents List tab, then open an agent.
On the Build tab, look at the Model menu at the top left. Write down the model for each agent. If it says GPT 4.1, that agent needs to move.
Part 2 · Switch and test
In the Model menu, choose GPT 6 Sol. Pick the one tagged Recommended. GPT 6.1 Sol is listed just below it.
Before you trust it with customers, click Test your agent (leave Starts with on Inbound) on the right side of the Build tab and ask it the five questions your customers ask most. Include one that should lead to a booking. You can test before you save. Test messages are never sent to your contacts.
If it sounds right, click Save. Save is only active after a change, so if you switched models it will be clickable. If the tone is off (too long, too formal, too pushy), adjust your agent's instructions and test again.
A real-world example
A roofing company's after-hours text agent
A roofer set up a texting agent last spring to answer storm-damage leads at night. It was on GPT 4.1. The owner switches it, tests the questions customers actually ask, and tightens the replies before saving.
Switch
The owner switches the agent to GPT 6 Sol.
Test
He texts it the usual: "Do you do insurance claims?", "How fast can you come out?", "How much for a roof inspection?"
Tighten
The answers are good, but a little longer than before, so he adds one line to the agent's instructions: "Keep replies to two sentences."
Save
He tests again and saves.
Tips from our team
- Use your real questions to test. Pull a few recent customer messages from Conversations and paste them in word for word.
- Watch the booking step most closely. That's where a change in tone costs you money.
- Switch one agent at a time. If you run several, you'll know which change caused what.
- Check again in a week. Read a few real conversations after the switch to make sure nothing drifted.
FAQ
Your quick win this week
Check one agent this week
Open Conversation AI, see which model the agent uses, and if it says GPT 4.1, switch it to GPT 6 Sol and test before you save. If you'd rather we do it for you, reply to our email.
Works best alongside
You change the model on each Conversation AI agent yourself.