A long context for little money
GigaChat 3.5 Ultra and 3 Pro take 262,144 tokens — eight times YandexGPT’s 32,768 — at $1.58 and $1.20 per million input tokens.
GigaChat by API, paid in US dollars
GigaChat models through an OpenAI-compatible API, on the same key as the rest of our catalog. Requests are processed in Russia; companies abroad pay by USD invoice from $50.
In short: GigaChat is not on OpenRouter as of October 2026, and a direct contract for the GigaChat API needs a Russian company or sole proprietor. Through kosareva.cloud you connect all GigaChat models with one key: only base_url and the key change, your OpenAI SDK code stays the same. Requests are processed in Russia. A company outside Russia pays by USD invoice from Kosareva Cloud LLC (Armenia), minimum $50.
Invoice in US dollars
Bank transfer to ID Bank, Armenia
OpenAI-compatible API
Works with any OpenAI SDK
Pay as you go
No subscription, billed per token
Seller named openly
Kosareva Cloud LLC, Armenia
/ Why GigaChat
Where GigaChat fits
Not a replacement for Western flagships in code and complex reasoning, but a long context at a low price for Russian text, processed in Russia.
Documents and correspondence in Russian
Long contracts, email threads and support history in Russian, answers from an internal knowledge base. A 30-page document fits into one request whole.
Processed in Russia
Requests to GigaChat are processed in Russia. A key marked “Russia-hosted models only” cannot reach a model hosted elsewhere, even by mistake.
Tool calling
The vendor declares tool calling for GigaChat 3.5 Ultra, 3 Pro and 3 Lightning: pass tools in the OpenAI format. For GigaChat 2 Max it is not declared.
The cheapest Russian chat model here
GigaChat 3 Lightning costs $0.20 per million tokens for both input and output, with the same 262,144-token context: labelling, field extraction, a first pass before a larger model.
Where GigaChat is weaker
The vendor adds a service prompt to every request, and you pay for it: about 1,214 tokens on 3 Pro and 990 on 3 Lightning. No cache discount, no backup route, and the maximum answer length is not published.
/ Catalog
GigaChat models in the catalog
Approximate prices in US dollars per 1M tokens: the catalog price in rubles divided by 85.5 ₽ per $1, ID Bank’s non-cash rate on 3 October 2026. Your balance is kept in rubles.
Sber262K context
GigaChat 3.5 Ultra
The flagship: 262K context, tool calling, a service prompt of only about 65 tokens.
$1.58 in · $4.73 out, per 1M tokens
Sber262K context
GigaChat 3 Pro
A quarter cheaper than Ultra per million, but adds about 1,214 service tokens to each request.
$1.20 in · $2.89 out, per 1M tokens
Sber262K context
GigaChat 3 Lightning
The smallest and cheapest GigaChat, for pipelines on texts of a few hundred tokens and up.
$0.20 in · $0.20 out, per 1M tokens
Sber131K context
GigaChat 2 Max
The previous generation, 131K context: kept for integrations already tuned to it.
$9.32 in · $9.32 out, per 1M tokens
/ Setup
First request to GigaChat
OpenAI-compatible endpoint: only base_url and the model name change, the code stays the same.
- /1
Register the company
Choose “Company or sole proprietor” → “In another country”: country, registration number, tax ID if you have one.
- /2
Top up by USD invoice
“Top up balance” → “Issue invoice (companies)”: an English invoice in US dollars, minimum $50, due in 10 days.
- /3
Send the first request
Create a key and set the model: gigachat-3-5-ultra, gigachat-3-pro, gigachat-3-lightning or gigachat-2-max. Usage, with the actual input tokens, shows up in the dashboard right away.
from openai import OpenAI client = OpenAI( api_key="YOUR_API_KEY", base_url="https://api.kosareva.cloud/v1", ) response = client.chat.completions.create( model="gigachat-3-5-ultra", messages=[{"role": "user", "content": "Summarize this contract clause in Russian in three sentences: ..."}], ) print(response.choices[0].message.content)
import OpenAI from "openai"; const client = new OpenAI({ apiKey: "YOUR_API_KEY", baseURL: "https://api.kosareva.cloud/v1", }); const response = await client.chat.completions.create({ model: "gigachat-3-5-ultra", messages: [{ role: "user", content: "Summarize this contract clause in Russian in three sentences: ..." }], }); console.log(response.choices[0].message.content);
curl https://api.kosareva.cloud/v1/chat/completions \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "gigachat-3-5-ultra", "messages": [{"role": "user", "content": "Summarize this contract clause in Russian in three sentences: ..."}] }'
/ Cost
What a month costs
Two typical workloads over 30 days, at the dollar prices above. Chatbot: 500 requests a day, 800 tokens in and 300 out. Documents: 100 requests a day, 20,000 tokens in and 800 out. The service prompt the vendor adds to each request is included.
| Model | Chatbot, per month | Documents, per month |
|---|---|---|
| GigaChat 3.5 Ultra | $41.71 | $106.18 |
| GigaChat 3 Pro | $49.12 | $83.03 |
| GigaChat 3 Lightning | $6.26 | $13.06 |
| GigaChat 2 Max | $155.36 | $582.03 |
Service prompt per request, measured on a “hi” on 20 September 2026: 65 tokens on 3.5 Ultra, 1,214 on 3 Pro, 990 on 3 Lightning, 11 on 2 Max. Approximate, at 85.5 ₽ per $1 (ID Bank, 3 October 2026). The balance is debited in rubles for actual tokens; there is no subscription.
/ Before you start
Who can buy, payment and data
Your contract is with us
Companies outside Russia register as “Company or sole proprietor” → “In another country”. Your contract and payment are with Kosareva Cloud LLC (Armenia), not with the model vendor. You are responsible for compliance with the rules of your jurisdiction; if in doubt, check with your counsel.
USD invoice from Armenia
Kosareva Cloud LLC issues the invoice in English and in US dollars to our USD account at ID Bank. The balance is credited after the bank confirms the payment, usually in 1–3 business days.
Data is processed in Russia
Requests to GigaChat are processed in Russia. If data protection rules apply to you, assess the transfer before you send personal data.
No foreign cards, no EUR yet
Cards issued outside Russia are not accepted. Invoices are in USD; for EUR, write to us and we’ll discuss it.
/ FAQ
Questions
The key, the price and choosing a model.
Where do I get a GigaChat API key outside Russia?
At kosareva.cloud the key is issued in the dashboard right after registration and works with all GigaChat models and the rest of the catalog. GigaChat is not on OpenRouter as of October 2026, and a direct contract for the GigaChat API needs a Russian company or sole proprietor.
Which GigaChat should I choose?
GigaChat 3.5 Ultra — for most work: the flagship, with a service prompt of only about 65 tokens. GigaChat 3 Pro — when requests run past about 3,500 tokens: then its lower price per million outweighs the 1,214-token service prompt. GigaChat 3 Lightning — for pipelines on texts of a few hundred tokens and up, at $0.20 per million. GigaChat 2 Max — only for integrations already tuned to it.
Why does the dashboard show more input tokens than I sent?
The vendor adds its own service prompt to every request, and it is billed as input: about 1,214 tokens on GigaChat 3 Pro, 990 on 3 Lightning, 65 on 3.5 Ultra and 11 on 2 Max (measured on 20 September 2026). On short requests this, not the price per million, decides the bill.
Can a company in the EU use GigaChat here?
Yes. You sign with and pay Kosareva Cloud LLC (Armenia) by USD invoice, not the model vendor. Requests are processed in Russia, so if GDPR applies to you, assess the transfer before sending personal data. Whether your own compliance rules allow it is for you to check, with your counsel if needed.
How does a company outside Russia pay?
By bank transfer against a USD invoice from Kosareva Cloud LLC (Armenia): minimum $50, due in 10 days. Your balance is kept in rubles; the invoice is converted at ID Bank’s non-cash rate on the invoice date. Cards issued outside Russia are not accepted.
Does data leave Russia?
Requests to GigaChat are processed in Russia. To make sure a key can’t reach a model hosted elsewhere, mark it “Russia-hosted models only” when you create it. If data protection rules apply to you, assess the transfer before sending personal data.
/ Next
Related pages
Count on your own requests
Create a key, send a dozen real requests to GigaChat 3.5 Ultra and 3 Pro and compare the actual input tokens in the dashboard: that, not the price list, shows which one is cheaper for you.