Notion AI handles the routine around documents: summarizing a long memo, answering from the knowledge base, pulling clauses out of a contract. Open models cover the same routine on your own server: internal documents never go to someone else's cloud, there is no volume cap, and the model can be fine-tuned on your templates or made to fill a JSON schema you define — client cards, fields from statements. Some of these models are built for document work rather than chat, and they do not demand heavy hardware. What you do not get is the editor, collaboration and access control — that is the product you build around the model. And a summary is still a retelling, so figures and deadlines from a contract are verified against the original.
Russian-language models for working with documents rather than chatting: knowledge-base answers, extraction of entities and facts from Russian text, long context.
Sber open models with strong Russian language support and local context, from 10B-A1.8B to 702B, all MIT. GigaChat3.1-Audio handles recordings up to two hours; GFusion is a fast diffusion text version.
A family of language models with strong Russian language support, from small versions for a laptop to a flagship on par with commercial APIs.
Models for template-based data extraction: give it a document or scan and a JSON field template, get a filled-in JSON back. NuExtract3 (4B) also converts scans to Markdown.
Compact Google models that run well on a single computer; larger versions understand images. Includes CodeGemma for code, FunctionGemma 270M for function calling and the fast DiffusionGemma.
Meno, GigaChat and NuExtract are listed as commercially usable; some Gemma versions fall under the Gemma Terms of Use and the larger Qwen models carry the developer's own license. For contract work, have the chosen version's license reviewed by a lawyer.
An open model can run on your own server: data stays in-house, there is no per-request fee, and the model can be fine-tuned on your documents.
