Running your models
We pick the right model for each job, run it where your rules allow, and keep it running.
Who it’s for
Companies that can't or don't want to send their data abroad. In Türkiye that includes every licensed financial firm, 483 of them, because their main systems must stay in the country. The KVKK guide on generative AI treats sending data to a foreign model as a transfer abroad.
How it runs
We test candidate models on your own tasks, in Turkish or any language where it matters, and compare quality, speed and cost across models and hosts. Then we run the winners where they belong: on your servers, on GPUs in Türkiye through our partners, or with a cloud provider you trust. When a better model comes out, we test it and switch if it earns it.
What you get
A short report on how each model did on your tasks, a running setup, and a monthly service that keeps it up and up to date.
Work we’ve done
We trained, evaluated and deployed open source Large Language Models on various hardware setups. We built an open-source tool for comparing models and hosts on the same tasks, and we use it for our own measurements of how today's models handle various scenarios.