ZeroTwo adds new models and retires old ones over time. If a chat was using a model that has been retired, ZeroTwo keeps the chat and sends your next message to Auto or to a current model instead of letting it fail. You can pick a different model at any time.
When a model is retired
- Your chat stays. ZeroTwo maps the retired model onto a live one for the chat's next message.
- Many retired models move to Auto. Release 1.1.29 lists Sonar, Sonar Pro, Kimi K2.5 and Mistral Small among them. Auto is ZeroTwo's own choice that routes to a fast, low-cost model. In Chat, Work and Code it routes to GPT 6 Luna.
- Some move to their successor. Older DeepSeek, Kimi and MiniMax models move to the current model rather than to Auto. Later retirements work the same way: Gemini 2.5 Pro moves to Gemini 3.1 Pro, Gemini 2.5 Flash to Gemini 3.1 Flash-Lite, GLM 5.2 to GLM 5.3, DeepSeek V4 Flash to DeepSeek V4.1 Flash, Mercury 2 and 2.5 to Mercury 3, and Mistral 3.1 24B to Gemma 4.
- Retired Grok models move to Auto. This covers the older Grok 4 and Grok 4.1 Fast models and Grok Code Fast.
- Models that only work on desktop. A model that needs the desktop app, such as one that runs on your own provider subscription, is replaced by Auto if the chat is opened on the web.
When your model is not on your plan
If the model you picked needs a paid plan or purchased credits you do not have, ZeroTwo sends your message with a model your plan covers instead of refusing it. A notice names both models, for example "… needs a paid plan or purchased credits. Switched to …", and the picker keeps the new choice.
Auto counts against your plan's free Auto requests. On the Free plan every Auto request uses credits. See How do credits work in ZeroTwo?
Choose a different model
- Open the model picker in the chat and choose the model you want.
- If the chat already has messages, ZeroTwo asks Switch models? and explains that switching mid-conversation uses additional credits, because the new model has to process the whole conversation. Choose Continue. Choose Continue and don’t show again to stop seeing the question.
To see how another model would have answered, open the retry menu under a response and choose Switch model.
What stays the same
Your personalization is not tied to a model. Memory, which ZeroTwo uses while Enable memory is on, and your custom instructions keep working when the model changes. You manage both in Settings > Personalization.
See what changed
Select the question-mark button (Help and what's new) in the sidebar footer. The What's new list shows recent releases, and Full changelog opens all of them. Release notes name the models that were added and say where chats on retired models went.