Models tested
Yandex, Sber, Qwen, MTS
Call modes
tool calling and standard prompting
Vendor lock-in
the model is a parameter, not a foundation
The model is your choice
Proven on live models
Not benchmarks — working dialogues. The platform has been tested with Russian cloud models and open ones — including a run inside a closed contour.
YandexGPT
Yandex’s cloud model — a fast start inside the Russian contour.
GigaChat
Sber’s model — tested on live dialogues of the platform.
Qwen
An open model — in the cloud or entirely inside your contour.
Cotype · MTS
The MTS corporate model — proven in production scenarios.
Your own LLM
A corporate model behind your API — the platform connects to it.
Switching by setting
Moving to another model is a configuration parameter, not a project.
One layer, any model
Dialogues, RAG and skills talk to the model layer, not to a particular vendor. Which is why a cloud LLM and a corporate one are equals to the platform.
A mode to match the model
Speed does not depend on the vendor
Streaming levels out the experience. Whichever model stands behind the layer, the answer starts typing into the dialogue in under a second.
First token
the answer starts typing into the dialogue at once
Models tested
Yandex, Sber, Qwen, MTS
Call modes
tool calling and standard prompting
Vendor lock-in
the model is a parameter, not a foundation
The cloud or your contour
Where the model lives is your decision. The platform works the same with a cloud API and with a model deployed inside the company — all the way to a fully isolated contour.
Connect your own model
Start in the cloud, move to your own The platform will not notice the difference: the knowledge base, the skills and the dialogues stay where they are.
Test it on your own data
Upload your own documents and catalogue — then ask the assistant the questions people ask you. That is the only test that means anything.