Language models

What language models are, how providers differ and how to connect the model you need to SKILLOK — for Space AI, tasks and agents.

About this section#

A language model is a program that continues text: it answers a question, writes code, summarizes an article, works through a problem. In SKILLOK a model works in Space AI, builds questions and drills from notes and acts as an agent in a repository — with your rights.

Here: how models work and what to expect from them, how to choose a model for the task and the budget, an article about each major provider — what it is good at, whether it works from Russia and how to connect its models in SKILLOK — and a catalog of models with prices.

Start here

What is a language model How a large language model (LLM) works, what tokens and context are, what a model does well and where it makes mistakes. How to choose a model Which model to take for the task — a fast and cheap one or the smartest — and how not to overpay. Model providers The companies that make language models — what their models are good at, whether they work from Russia and how to connect them in SKILLOK.

In this section

What is a language model
How a large language model (LLM) works, what tokens and context are, what a model does well and where it makes mistakes.
How to choose a model
Which model to take for the task — a fast and cheap one or the smartest — and how not to overpay.
How to connect a model in SKILLOK
Three ways to give SKILLOK a model — the SKILLOK model, your own OpenRouter key or your own DeepSeek key — and how to use them in Space AI.
Model providers
The companies that make language models — what their models are good at, whether they work from Russia and how to connect them in SKILLOK.
Language model catalog: prices and capabilities
37 current models from OpenAI, Anthropic, Google, DeepSeek, Qwen, xAI, Moonshot, Z.ai, Mistral and Meta — price per million tokens, context window, what the model accepts and how to connect it in SKILLOK.