Every top AI model in one place. One key. No subscription.
−79%Same models, up to 79% below official prices
GPT, Claude, Gemini, DeepSeek, Grok and dozens more models. Top up once and use any of them.
- SBP, card or crypto
- OpenAI and Anthropic API
- A spending limit for each key
- We don't store your prompts
- Claude Opus 5.5$2.40
- GPT-6.1 Sol$0.80
- Gemini 4 Argon$2.00
- Grok 4.7$0.42
- DeepSeek V4 Pro$0.396
- Kimi K3.1 Pro$1.20
- Qwen3.8 Max$1.73
- GLM-5.3$0.93
- MiniMax M3.1 Flash$0.15
- Doubao Seed 2.1 Pro$0.69
- LongCat 2.0$0.30
- HY4$0.834
- Claude Sonnet 5.5$1.20
- GPT-6.1 Terra$0.40
- Gemini 3.8 Flash$0.75
- Grok 4.6$0.42
- Kimi K3.1 Pro Highspeed$4.00
- Qwen3.7 Plus$0.2875
- GLM-5.3 Flash$0.15
- MiniMax M3$0.45
- Doubao Seed 2.1 Lite$0.115
- Claude Fable 5.1$9.56
- GPT-6 Astra$3.00
- Grok 4.5$0.42
- Kimi K3$2.40
- Doubao Seed 2.0 Mini$0.115
- Claude Fable 5$9.56
- GPT-6 Sol$0.80
- Grok 4.3$0.75
- Kimi for Coding$1.00
- Claude Opus 5$3.00
- GPT-5.6 Sol$1.60
- Claude Sonnet 5$1.20
- GPT-5.6 Terra$0.80
- Claude Opus 4.8$3.00
- GPT-5.6 Luna$0.20
- Claude Opus 4.7$3.00
- Claude Opus 4.6$3.00
- Claude Sonnet 4.6$1.80
- Claude Haiku 4.5$0.80
One API instead of a dozen SDKs, accounts and subscriptions. Switch models, not code. Pay only for what you use.
42 models from the leading AI labs
- Anthropic
- OpenAI
- xAI
- DeepSeek
- Moonshot AI
- Alibaba
- Zhipu AI
- MiniMax
- ByteDance
- Meituan
For everyone and every task
Code, writing, bots, automation or just chatting with AI: FreshRouter plugs into any tool that works with OpenAI or Anthropic.
Coding agents
Hit your plan's limit? Keep coding without waiting, on any top model.
- Claude Code
- Codex CLI
- Cursor
- Cline
- Roo Code
- +5 more
Apps and Telegram bots
Build on any model and switch it without rewriting code. Each service gets its own key and limit.
- OpenAI SDK
- Anthropic SDK
- Vercel AI SDK
- LangChain
- LlamaIndex
- +3 more
Everyday AI chat
Every top model in one chat app, with one balance instead of a subscription for each.
- Open WebUI
- LibreChat
- LobeChat
- Cherry Studio
- Chatbox
- TypingMind
Automation
Any model in your workflows, through their OpenAI-compatible settings.
- n8n
- Dify
- Flowise
- Langflow
Plug FreshRouter into anything in a minute
Three steps, or let your AI agent do them for you.
- 01
Create an account
Email or Telegram, about 30 seconds.
- 02
Top up your balance
SBP, card or crypto. Spend it on any model.
- 03
Connect your tool
Give the ready-made prompt to your agent, run one command, or paste our base URL and your key.
Paste it into Claude Code, Codex, Cursor or any other coding agent. It connects the tool it runs in, or your project if you say so.
Connect FreshRouter for me. It is one API key for Claude, GPT, Gemini, Grok, DeepSeek and other models, in the OpenAI and Anthropic API formats. - My API key: ask me for it first (it starts with sk-), then use it below in place of <key>. - OpenAI-compatible base URL: https://freshrouter.com/v1 (chat completions and the Responses API) - Anthropic-compatible base URL: https://freshrouter.com (no /v1 at the end; for Claude models) - Every model with its ID, context size and price: https://freshrouter.com/models.json (no key needed) - A guide for each tool, written for agents: https://freshrouter.com/llms.txt What to connect: the tool you are running in. If I add something after this prompt, such as another tool's name or "this project", connect that instead. 1. Claude Code, Codex, Grok Build and Pi have installers. Run the command for the tool, and nothing else is needed: Claude Code: curl -fsSL https://freshrouter.com/install/claude.sh | bash -s -- <key> Codex: curl -fsSL https://freshrouter.com/install/codex.sh | bash -s -- <key> Grok Build: curl -fsSL https://freshrouter.com/install/grok.js | node - <key> Pi: curl -fsSL https://freshrouter.com/install/pi.js | node - <key> In PowerShell on Windows: Claude Code: $env:FRESHROUTER_API_KEY="<key>"; irm https://freshrouter.com/install/claude.ps1 | iex Codex: $env:FRESHROUTER_API_KEY="<key>"; irm https://freshrouter.com/install/codex.ps1 | iex Grok Build: irm https://freshrouter.com/install/grok.js | node - <key> Pi: irm https://freshrouter.com/install/pi.js | node - <key> Each one backs up the file it changes and is safe to run again. 2. Any other tool (Cursor, Cline, OpenCode, OpenClaw, Aider, Continue, Zed and so on): add FreshRouter to the tool's own settings as a custom OpenAI-compatible provider, with the base URL and key above. Look up how in the tool's own docs or its --help first; don't guess the format. If the tool wants a list of models, take the IDs and context sizes from the model catalog. Keep the rest of its settings and back up every file you change. If it can only be set in the tool's app, tell me exactly what to paste where. 3. "This project": change only the base URL and key of the OpenAI or Anthropic SDK the project already uses, or add the official OpenAI SDK for the project's language. Put the key in the project's .env file as FRESHROUTER_API_KEY, make sure .env is in .gitignore, and never hard-code or commit the key. Keep the model ID in one setting. Don't ask me which model to use: start on claude-sonnet-5-5, or on gpt-6.1-sol where only OpenAI models fit, and I'll change it later. When you're done, tell me what you changed and how to start using it. For a project, also send one short test request and show me the reply.
Clear prices, no catches
Prices per 1M tokens, in USD, with no hidden fees.
Anthropic
40% below the official priceClaude Opus 5.5
Anthropic's top model, strong at code and agents
$2.40in
$12.00out
OpenAI
70% below the official priceGPT-6 Astra
OpenAI's most capable model, for the hardest tasks
$3.00in
$15.00out
xAI
79% below the official priceGrok 4.7
xAI's newest model, with a low output price
$0.42in
$1.26out
Anthropic
40% below the official priceClaude Sonnet 5.5
Fast, capable Claude for everyday work
$1.20in
$6.00out
OpenAI
60% below the official priceGPT-6.1 Sol
OpenAI's everyday workhorse for code and text
$0.80in
$4.00out
OpenAI
60% below the official priceGPT-6.1 Terra
Good quality for little money: bots and big volumes
$0.40in
$2.00out
In is what you send, out is what the model writes back. Discounts are against each provider's own official price as of November 17, 2026.
Real models. Honest bills.
Six things you can hold us to.
The model you name is the one that answers
Your request goes to the model you asked for, unchanged. We never swap in a cheaper one.
Repeated context costs many times less
Coding agents send the same files again on every step. We bill that part at the cached price.
Every request is on the record
Time, model, tokens in and out, and what it cost: each one is listed in your console.
The rate is fixed when you pay
You top up in rubles. You see the rate before you pay, and it holds for that payment.
A failed request costs nothing
If a request fails before the model answers, nothing is taken from your balance.
Your key stays yours
It's shown once, can have its own spending limit, and your account can ask for a second step at login.
Questions and answers
- What is FreshRouter?
- One API key for dozens of top AI models. Connect once, top up, and use any model, paying only for the tokens you spend.
- How does billing work?
- You top up a prepaid balance with SBP, a card or crypto. Each request is charged for the tokens it uses. No subscription, no monthly fee.
- Can I limit what a key can spend?
- Yes. Each key can have its own spending limit, list of allowed models, IP allowlist and expiry date.
- Do you store my prompts?
- No. Requests only pass through us on their way to the model's provider. We keep just the time, model, token count and cost, and you see them in your console.
Start building in minutes
Create a free account, top up your balance and connect your first tool, bot or app.
Create free account