Free is funded by free open models served by third-party hosts. The cost is not money; it is that those hosts set their own data rules.
The difference
| Free | Plus and Pro | |
|---|---|---|
| Models | Free open models only | Premium and open models |
| Host logging | Hosts may log prompts | OpenRouter requests restricted to endpoints that say they do not collect or train on prompts |
| Training by hosts | Some hosts may use prompts to improve models | Direct OpenAI and NVIDIA calls are handled under their API terms |
| Keplar training | Does not train on your data | Does not train on your data |
What this means in practice
- A Free question may be seen by a host Keplar does not control.
- You cannot ask a host to forget it through Keplar.
- Deleting a chat removes it from Keplar. It cannot remove data a provider keeps under its own terms.
What to keep out of Free
Passwords, keys, account numbers, health or legal details about you or others, unpublished work you do not want a third party to see, and anything you promised to keep confidential. If it is sensitive, do not ask it on Free.
Why Keplar is open about it
Because it is a trade-off, not a flaw. The Free screen and pricing tiers state it, and so does the Privacy Policy. Keplar Free never routes you to a paid model without telling you.
If you need more
A paid plan sends requests only to endpoints that state they do not collect or train on prompts, though providers may retain requests for a limited time. Business customers who need contractual terms can use the Inquire form.
Related
- Free plan limits and behavior: Up to 50 questions a day, free open models only, the three-model cap, stand-ins when models are busy, refunded daily counts and what Free cannot do.
- How Keplar handles your data: Where a question goes, what is stored under your account, which companies process it, what is never sent to model providers, and what Keplar does not do.