Claude is usually brought in for long document analysis, step-by-step reasoning and agent work with tools. In the open world those jobs are covered by reasoning models and large mixture-of-experts families: they deploy inside your perimeter, contracts and correspondence stay in house, there is no per-request billing, and the behaviour can be tuned to your own policies. The trade-off is hardware: reasoning answers are longer than ordinary ones, and the larger versions of these families expect server-grade GPUs, sometimes several at once. On tangled, long-context tasks an open model errs more often than a flagship cloud service, so conclusions about contracts and numbers still need a human check. A sensible start is a mid-size model tested on your own documents, with the larger tier considered only afterwards.
A family of language models with strong Russian language support, from small versions for a laptop to a flagship on par with commercial APIs.
A reasoning model that thinks step by step before answering. Strong at calculations, logic and code; compact distilled versions are available.
OpenAI's first open models since GPT-2. Reasoning and tool calling; the smaller version fits on a single GPU.
Very large Moonshot MoE models for agentic work. K3 (2.8 trillion parameters) was the largest open model at release, with up to 1M tokens of context and image understanding; K2.7-Code is built for programming.
One of the oldest Chinese open lines: from ChatGLM-6B to GLM-5.3. Strong at agentic tasks and programming; GLM-5.3-Flash understands images and is released under MIT.
DeepSeek-R1 and gpt-oss ship under MIT and Apache 2.0, while some of the larger Kimi and GLM versions carry their own licenses with carve-outs for large companies and API providers. If you plan to resell access to the model, review those terms separately.
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.
