Only original source-side text characters are counted, the translated target-side text is not counted.
One character is counted the same regardless of whether it is a whitespace character, a decomposed umlaut or a UTF-8 Chinese ideogram. This avoids lossy conversions and makes it easy for client applications to know the exact count.
You are only billed for successful responses, not for requests that fail due to bad requests or server errors.
$200 per 1M characters is $0.0002 per character or about $0.001 per word. (We assume 5 characters per word.)
Yes! Right now new accounts get free credit for 1M characters in the console and 100K characters for the API.
We also support non-profit research and humanitarian projects with additional free credits.
Both plans include the API, evaluations and console and all languages.
By default, the API is available with low latency and high availability in data centers around the world, similar to the major machine translation APIs. As we keep adding languages and error types, rolling out new datacenter locations and building new console features, they will be available to all clients. We're constantly improving our accuracy on different topics and types of texts, and those improvements get rolled out to clients on the Standard plan too.
Bottom line: no crippleware! The partner plan is available to support higher scale and extra investments in custom development.
Training a custom model or other customization requires a minimum commitment, according to the level of customization.
Deploying a custom model requires a minimum commitment of $1K per month.
Yes! We offer discounts for volumes above 100M characters per month and for large offline batch jobs. We also offer free additional characters to clients who provide significant amounts of relevant high-quality data or feedback.
Yes! Under the Standard plan, you can cap your total usage to avoid surprises, and only pay for what you use. Under the Partner plan, there is additional flexibility.
Yes! Our Standard plan terms do not mention any attribution requirements, nor any blanket restriction against including ModelFront results in other platforms as a paid, free or bundled feature.
If there is attribution - for example, if the ModelFront name or logo are used - then we should work together to make sure that it reasonably reflects the accuracy and performance of ModelFront and is understandable to users.
No. We work with clients and partners around the world, and taxes are treated differently in each country.
For questions and answers about translation risk prediction technology, read about our approach. For questions about and our API and console, see our general FAQ, documentation and one-pager.
You can also read about ModelFront's default security, data privacy and confidentiality, terms of service and service-level agreement.
Have a special use case? Want more info?
We're here to listen, answer your questions and offer the plan that's right for you.Let's talk!