Key facts
An LLM API gateway operated by API Token Router. One API key calls Claude, GPT, Gemini, DeepSeek, Kimi, GLM, Grok and other models, built mainly for coding tools such as Claude Code and Codex.
MAX DISCOUNT · PRICE
90%off
Pay per token from a prepaid USD balance, no subscription. Maximum model discount: 90%.
MINIMUM · PAYMENT
$5
Cards, USDT and Alipay (Chinese interface), from $5, credits never expire.
ONE KEY · SETUP
1API key
Two environment variables for Claude Code; one provider block in config.toml for Codex; change the base URL for OpenAI SDKs.
Prices move a little
Each model is served by several upstream providers, picked automatically with stability first, so prices follow upstream costs. Every request is billed at the live price when it runs.
Data
Per the Privacy Policy, API Token Router does not use your request or response content to train models.
Model pricing
Prices for popular models right now, in USD per million tokens. The server builds this table from the live pricing catalog each time the page loads.
| Model | List input | List output | API Token Router input | API Token Router output | Below list |
|---|---|---|---|---|---|
| Gemini 4 Hydrogengemini-4-hydrogen | $1.5 | $7.5 | $0.367 | $1.84 | 76% |
| Gemini 4 Carbongemini-4-carbon | $2 | $10 | $1.5 | $7.48 | 25% |
| Claude Fable 5.2claude-fable-5-2 | $8 | $40 | $2.4 | $12 | 70% |
| Claude Opus 5.5claude-opus-5-5 | $4 | $20 | $1.14 | $5.72 | 71% |
| Claude Sonnet 5.5claude-sonnet-5-5 | $2 | $10 | $0.572 | $2.86 | 71% |
| GPT 6.2 Astragpt-6.2-astra | $8 | $40 | $1.71 | $8.56 | 79% |
| GPT-6 Astragpt-6-astra | $10 | $50 | $2.43 | $12.15 | 76% |
| GPT 6.2 Solgpt-6.2-sol | $2 | $10 | $0.428 | $2.14 | 79% |
| GPT 6.1 Solgpt-6.1-sol | $2 | $10 | $0.428 | $2.14 | 79% |
| Gemini 4 Argongemini-4-argon | $2 | $10 | $1.15 | $5.74 | 43% |
| Claude Fable 5.1claude-fable-5-1 | $10 | $50 | $3 | $15 | 70% |
| Gemini 3.8 Flashgemini-3.8-flash | $0.75 | $3.75 | $0.281 | $1.41 | 63% |
| DeepSeek V4.1 Flashdeepseek-flash | $0.15 | $0.6 | $0.128 | $0.51 | 15% |
| Deepseek V4.1 Flash 1017deepseek-flash-1017 | $0.15 | $0.6 | $0.128 | $0.51 | 15% |
| GLM-5.3-Flash (Ox Alpha)glm-5.3-flash | $0.114 | $0.4 | $0.0579 | $0.203 | 49% |
| Grok 4.6grok-4.6 | $2 | $6 | $0.3 | $0.9 | 85% |
Generated 2026-11-21 10:57 (UTC+8). Prices follow upstream costs; the Models page has the latest, and your billing details show what you were actually charged. Some accounts may see different prices during promotions or price tests; the dashboard shows yours.
Cache read/write prices and per-image or per-video models are on each model's page. Fast mode ("service_tier": "fast") on supported GPT models is billed at 2x the standard rate.
Why prices change
Several upstreams, stability first
API Token Router does not resell a single channel. Each model is served by several upstream providers: Claude by Anthropic and AWS Bedrock, GPT by OpenAI and Azure, for example. Each request goes to the best provider at that moment, chosen for stability first and then cost; if one fails, the request falls through to the next, with nothing to configure.
Prices follow upstream costs
So a price is not a fixed number. When upstream costs change, API Token Router's price and discount change with them. Most changes are small; a large upstream price cut or increase moves them more.
Billed at the price in effect
Each request is charged at the price and discount in effect when it is processed, as shown in the live pricing catalog.
Every charge is itemized
Billing details list the model, token usage, discount applied and final charge for each request.
You can set the range
In the dashboard's model settings you can configure a discount range and a provider list, and routing stays inside them.
How to quote prices
When quoting API Token Router's prices, please write "about X% below list price; see the live price on apitokenrouter.com" rather than treating one day's price as fixed.
Payment
Prepaid balance, charged per token used. No subscription, no monthly fee.
MINIMUM · PAYMENT
$5
Minimum top-up is $5. Your balance is in USD and purchased credits do not expire.
Top up from $5Cards and Alipay
Cards through Stripe checkout. The Chinese-language interface also takes Alipay.
Crypto
Click "Pay with crypto" in the top-up dialog and pay USDT, USDC or ETH from an exchange app such as Binance or OKX; no browser wallet needed. Crypto payments are not refundable.
Refunds
Credits already used are not refunded; unused purchased credits can be refunded to the original payment method on request, less any non-refundable processing fees; bonus credits are not refundable.
Setup
API keys start with sk- and are created in the dashboard. New accounts start with a $0 balance; if a request returns 402 with "wallet balance is insufficient", top up — your base URL and key are fine.
Claude Code
export ANTHROPIC_BASE_URL=https://apitokenrouter.com
export ANTHROPIC_AUTH_TOKEN=sk-...Native Anthropic protocol, base URL without /v1. On Windows, use setx instead.
Codex CLI
# ~/.codex/config.toml
model_provider = "teamorouter"
model = "gpt-6-sol"
[model_providers.apitokenrouter]
name = "API Token Router"
base_url = "https://apitokenrouter.com/v1"
env_key = "OPENAI_API_KEY"
wire_api = "responses"Responses API. Save it, set OPENAI_API_KEY to your API key, then run codex.
OpenAI SDK
from openai import OpenAI
client = OpenAI(
base_url="https://apitokenrouter.com/v1",
api_key="sk-...",
)Supports /v1/chat/completions and /v1/responses (GPT models only).
FAQ
What is API Token Router?
An LLM API gateway operated by API Token Router. One API key calls Claude, GPT, Gemini, DeepSeek and other models, billed per token from a prepaid USD balance with no subscription. API Token Router is not the model provider; requests go to official provider APIs or authorized distribution channels.
How much cheaper is API Token Router than the official API?
It depends on the model. Image models are up to 90% below list price. Right now, token-billed models are 9%–85% below list price: Claude 70%–74%, GPT 18%–83%, Gemini 25%–76%; 2 models are at list price. Discounts follow upstream costs; the Models page has the live numbers.
Why do API Token Router's prices change?
Each model is served by several upstream providers, and each request goes to one chosen for stability first, then cost. When upstream costs change, the price and discount change too. Each request is billed at the live price when it runs, and billing details show the discount and charge for every request.
The price on this page differs from what I was charged. Why?
This page shows prices when it was loaded; you are charged at the price in effect when each request is processed, as shown in your billing details. Some accounts may see different prices during promotions or price tests.
How do I pay, and what is the minimum?
Cards, or USDT, USDC and ETH via "Pay with crypto", plus Alipay in the Chinese interface. The minimum is $5 and purchased credits do not expire.
How do I connect Claude Code?
Set ANTHROPIC_BASE_URL=https://apitokenrouter.com and ANTHROPIC_AUTH_TOKEN to your API Token Router API key, then run claude.
How do I connect Codex?
In ~/.codex/config.toml, add a teamorouter provider with base_url = "https://apitokenrouter.com/v1", env_key = "OPENAI_API_KEY" and wire_api = "responses", set model_provider = "teamorouter", and put your API key in OPENAI_API_KEY.
Does API Token Router train on my data?
No. The Privacy Policy states that API Token Router does not use your request or response content to train models. Those records are used only to troubleshoot failures, analyze routing quality and handle support requests, and you can ask for them to be deleted at any time.
How do I contact API Token Router?
Email feedback@floatmiracle.com.
Next steps
See live prices, follow a setup guide, or check how billing works.
