Which AI models does Pivra use?
The Fast, Balanced and Most capable models, which plans include them, what changed in October 2026, and where the Creativity setting applies.
Updated 8 Oct 2026
Pivra's chatbots run on OpenAI models. Across the product we call them by what they're for: Fast, Balanced and Most capable. This page is the one place that names the model behind each, so it's the page to check when a model changes.
The models
| Tier | Model | Plans | Creativity setting |
|---|---|---|---|
| Fast | GPT-6 Luna | Free, Starter, Pro and Agency. New chatbots start here. | Used |
| Balanced | GPT-5.4 Mini | Pro, Business and Agency. Founding and Business chatbots always run on it. | Used |
| Most capable | GPT-6.1 Sol | Not offered on self-serve plans | Not used |
Free and Starter chatbots run on the Fast model. On Pro and Agency you choose Fast or Balanced for each chatbot in Settings → General → AI Model.
Your knowledge sources are searched with OpenAI's text-embedding-3-small model on every plan.
Pivra sends the text a reply needs to OpenAI's API. Our privacy policy explains what is sent and how it's handled.
What changed in October 2026
OpenAI deprecated GPT-5.4 Nano, our previous Fast model, on 1 October 2026 and will switch it off on 1 April 2027. An older model, GPT-4.1 Nano, switches off on 23 October 2026. So we moved to current models:
- Fast moved from GPT-5.4 Nano to GPT-6 Luna.
- Balanced is unchanged: GPT-5.4 Mini.
- Most capable moved from GPT-5.5 to GPT-6.1 Sol. It isn't offered on self-serve plans.
These older models can no longer be selected: GPT-5.4 Nano, GPT-4.1 Nano, GPT-4.1 Mini, GPT-4.1, GPT-4o mini, GPT-4o and GPT-5.5.
Do I need to do anything?
No. A chatbot saved on an older model now runs on the current model for its tier, automatically. Your plan, price, message limits and knowledge sources stay the same, and you don't need to re-train anything.
If a chatbot was on an older model, Settings → General → AI Model shows a short notice. Saving the settings records the current model.
A new model can word things a little differently. It's worth asking your chatbot a few of your most common questions. For answers that must never vary, use refinements.
When the Creativity setting applies
Creativity (Temperature) in Settings → General → Response Tuning applies on the Fast and Balanced models, which covers every self-serve plan. Lower values keep answers consistent; higher values vary the wording more.
The Most capable model doesn't accept the setting, so Pivra doesn't send it there. It isn't offered on self-serve plans. See Improving chatbot responses.
Was this helpful?
If you have questions or suggestions, email us at support@pivra.ai .