OpenAI models in Yenisei: open GPT-OSS and migrating from the OpenAI API

The Yenisei catalog includes OpenAI models with open weights: GPT-OSS 120B and GPT-OSS 20B. We do not offer OpenAI's closed models: the GPT-5.6 lineup, GPT-5.5 Pro, o3 and o4-mini, or the text-embedding-3 embeddings. We work with open-weight models and our own Yenisei lineup. Below is what we have, what to use instead of the rest, and how to migrate.

Payment is in rubles: 1 credit = 1 ₽, charged for the tokens actually processed. Prices are given in rubles per 1M tokens as of the publication date; current prices are on each model's page.

Open GPT-OSS

• GPT-OSS 120B: 2.96 / 13.6 ₽, reasoning and code.
• GPT-OSS 20B: 2.4 / 10.4 ₽, the compact option.

With open-weight models, behavior and cost stay predictable at high request volumes.

What to use instead of OpenAI's closed models

• Instead of the top-tier GPT-5.6 and GPT-5.5 Pro: DeepSeek V4 Pro, Kimi K3 or our Yenisei Max.
• Instead of the fast and inexpensive GPT-5 Mini and Nano: GPT-OSS 20B, DeepSeek V4 Flash or Mistral Nemo.
• Instead of the reasoning models o3 and o4-mini: GPT-OSS 120B or Kimi K2 Thinking.
• Instead of the text-embedding-3 embeddings: BGE-M3 or Qwen3 Embedding 8B: both are multilingual and work well with Russian.

How to migrate

client = OpenAI(
    base_url="https://api.yenisei.ru/v1",   # was api.openai.com
    api_key=YENISEI_API_KEY,
)

After that, the code does not change: the same calls and SDK, streaming responses and tool calling work as before. The only thing to change is the model name in the model field. If you are migrating a search over your database, the vectors will have to be recomputed with the new embedding model: vectors from different models are not comparable. Keys, limits and usage are in the console.