About OpenKey

OpenKey is a unified API for large language models: one key, one OpenAI-compatible endpoint, every major model — currently 338 of them from 52 labs.

Why 3%

Model routing is a commodity. Switching between gateways is two lines of code, which means the only durable reason to stay with one is that you trust it. So we made the pricing impossible to misread: you pay the provider’s list price plus a flat 3%. The math is printed on every model page, every comparison page, and in every API response (usage.cost_usd).

That’s the whole model. No credit-purchase fees, no minimums, no spread between the price we quote and the price we bill, no “N/A” in the fee row.

What we hold ourselves to

• Honest errors. Infrastructure failures return 5xx and say what broke. The contract is written down in error semantics.
• A status page you can find. /status is public, indexable, and keeps its incident history.
• Rankings you can’t buy. Usage rankings are whale-filtered and the methodology is published. A single customer routing trillions of promo-priced tokens will not top our charts.
• Disclosure over vibes. Model pages show which providers serve each model and log version changes in the changelog, because “same model id” doesn’t always mean “same weights”.

Contact

Questions, incidents, or corrections: hello@openkey.ai. If a number anywhere on this site is wrong, we want to know within the hour.