AI for programming in Russia: Claude, GPT, DeepSeek and Qwen in your editor without a VPN
Which AI is best for programming depends on the task: long agent sessions over a project, quick edits and bulk simple jobs each suit different models. Through RuAPI they all work with one key: Claude Opus 5.5 and Sonnet 5.5, GPT-6.1, DeepSeek, Qwen, Kimi and GLM connect to Claude Code, Cline, Kilo Code, OpenCode, Codex and other tools from Russia without a VPN. You pay in USDT, only for the tokens you use.
export ANTHROPIC_BASE_URL=https://www.ruapi.ai
export ANTHROPIC_AUTH_TOKEN=sk-...
export ANTHROPIC_MODEL=claude-sonnet-5-5
# cheaper: export ANTHROPIC_MODEL=deepseek-flash
claudeThis is how Claude Code connects to RuAPI. Switch models with /model: the list includes Claude and models from other vendors that your key can use.
Why AI coding models are hard to get in Russia
Strong coding models are not available in Russia directly.
Anthropic and OpenAI do not offer their APIs in Russia: it is missing from their supported-country lists (anthropic.com/supported-countries, developers.openai.com, checked 10 Nov 2026). AI editor subscriptions are paid with a foreign bank card. A RuAPI key is issued right after you sign up with email, Google, Yandex or Telegram, and you top up in USDT.
There is no single “best” model.
A model that handles a multi-hour agent task well costs more than you need for autocomplete or renaming variables. It is easier to keep two or three models at hand and switch between them. In RuAPI they all work with one key and one balance, and the console log shows what each request cost.
Agents burn tokens faster than you think.
A coding agent resends the conversation history and the files it has read on every step. On a long task that adds up to hundreds of thousands of tokens. That is why prompt caching and the cache-read price matter: for claude-sonnet-5-5 it is $0.12 per 1M tokens versus $2.40 for new input. Give the agent’s key a budget in the RuAPI console.
How to start: 3 steps
Works with any tool that speaks the OpenAI-compatible API or the Anthropic API.
Sign up for RuAPI and top up
Sign up at www.ruapi.ai with email, Google, Yandex or Telegram. The minimum top-up is 5 USDT on TRC20, BEP20, TON or Polygon; USDC is accepted too. How to buy USDT for rubles via SBP is shown at docs.ruapi.ai/en/buy-usdt. Foreign bank cards also work.
A new account usually gets $0.05 right away — enough to try deepseek-flash, claude-haiku-5-5 or gpt-6-luna. The other models unlock after your first top-up.
Create a key and connect your tool
Console → “API keys” → “Create key”. In the “Connect an app” window pick your tool: Claude Code, OpenCode and Codex have a one-command install, the rest get ready-made settings. The address for OpenAI-compatible apps is https://www.ruapi.ai/v1; for Claude Code and other Anthropic-API apps it is https://www.ruapi.ai.
Pick a model for the task
- Long agent work on a large project, hard refactoring: claude-opus-5-5, or claude-fable-5-2 for the hardest tasks.
- Everyday development — edits, tests, debugging: claude-sonnet-5-5 or gpt-6.1-sol.
- Bulk and simple jobs, drafts, autocomplete: deepseek-flash, gpt-6-luna or qwen3.8-flash.
- When price matters on hard tasks: deepseek-v4-pro, qwen3.8-max, kimi-k3.1-pro and glm-5.3 cost less than Claude Opus.
RuAPI addresses and keys: docs.ruapi.ai/en/quickstart · prompt caching: docs.ruapi.ai/en/prompt-caching
What to enter in your tool
OpenAI-compatible apps
Base URL: https://www.ruapi.ai/v1
API Key: sk-…
Model: claude-sonnet-5-5, deepseek-flash…
Claude Code and Anthropic-API apps
ANTHROPIC_BASE_URL: https://www.ruapi.ai
ANTHROPIC_AUTH_TOKEN: sk-…
no /v1 at the end
Payment
USDT / USDC from 5 USDT
foreign bank cards
no subscription
Prices of coding models
In US dollars per 1M tokens, as on the pricing page. No subscription: only the tokens your requests use are deducted.
| model | input $/M | output $/M |
|---|---|---|
| claude-fable-5-2 | $9.60 cache read $0.24 | $48.00 |
| claude-opus-5-5 | $4.80 cache read $0.24 | $24.00 |
| claude-sonnet-5-5 | $2.40 cache read $0.12 | $12.00 |
| gpt-6.1-sol | $2.40 cache read $0.12 >272K: $4.80 | $12.00 >272K: $18.00 |
| deepseek-v4-pro | $1.58 cache read $0.0600 | $4.75 |
| qwen3.8-max | $2.16 cache read $0.43 | $6.46 |
| kimi-k3.1-pro | $2.40 cache read $0.24 | $12.00 |
| deepseek-flash | $0.17 cache read $0.0120 | $1.44 |
Input is the price of new input tokens, “cache read” of input tokens served from cache, output of the model’s reply. Coding agents repeat the start of the request on every step, so most of the input is usually read from cache. By our estimate, an agent step with 50K cached context tokens and a 1K-token reply costs about $0.018 on claude-sonnet-5-5 and about $0.002 on deepseek-flash. Your account’s prices may be lower.
The grey “>272K” row is the vendor’s long-prompt rate: if the prompt is over the threshold, the whole request is billed at it.
AI editor subscription or pay-per-token API
| Your own tool + RuAPI | AI editor subscription | |
|---|---|---|
| How to pay from Russia | USDT from 5 USDT, or a foreign bank card | Usually a foreign bank card only |
| What you pay | Only for tokens: almost nothing in a quiet month | A fixed monthly fee, extra above the limit |
| Choice of models | Any RuAPI model, switch without changing tools | The models the editor’s developer picked |
| Tools | Claude Code, Cline, Kilo Code, OpenCode, Codex, Aider, Zed and more | That editor only |
This compares payment models in general, not specific products: subscription terms differ between editors. A subscription pays off if you hit its limit every day; paying per token is better when your load is uneven.