The LLM gateway built for agents
One API key, one prepaid balance — every major model. Prices are USD per 1M tokens and track upstream costs: they refresh daily, and every request settles at the live rate at the moment it runs.
Prices refresh daily · Last checked
23
models behind one key
−87%
vs list price, at best
3
native protocols (Anthropic · OpenAI · Gemini)
$0
free tier — 50 requests/day
| Model | Context | Input (list) | Output (list) | Input (Cradler) | Output (Cradler) |
|---|---|---|---|---|---|
GPT-5.6 Sol gpt-5.6-sol | 1M | $5 cache $0.5 | $30 | $0.65−87% cache $0.065 | $3.9−87% |
GPT-5.6 Terra gpt-5.6-terra | 1M | $2 cache $0.2 | $12 | $0.26−87% cache $0.026 | $1.56−87% |
GPT-5.6 Luna gpt-5.6-luna | 1M | $0.2 cache $0.02 | $1.2 | $0.026−87% cache $0.0026 | $0.156−87% |
GPT-5.5 gpt-5.5 | 1M | $5 cache $0.5 | $30 | $0.65−87% cache $0.065 | $3.9−87% |
GPT-5.4 gpt-5.4 | 1M | $2.5 cache $0.25 | $15 | $0.325−87% cache $0.0325 | $1.95−87% |
GPT-5.4 Mini gpt-5.4-mini | 1M | $0.75 cache $0.075 | $4.5 | $0.0975−87% cache $0.0097 | $0.585−87% |
GPT Image 2 gpt-image-2 | — | $5 cache $1.25 | $30 | $1.25−75% cache $0.3125 | $7.5−75% |
Claude Opus 4.8 claude-opus-4-8 | 1M | $5 cache $0.5 | $25 | $1.95−61% cache $0.195 | $9.75−61% |
Claude Opus 5 claude-opus-5 | 1M | $5 cache $0.5 | $25 | $1.95−61% cache $0.195 | $9.75−61% |
Claude Opus 4.7 claude-opus-4-7 | 1M | $5 cache $0.5 | $25 | $1.95−61% cache $0.195 | $9.75−61% |
Claude Opus 4.6 claude-opus-4-6 | 1M | $5 cache $0.5 | $25 | $1.95−61% cache $0.195 | $9.75−61% |
Claude Sonnet 4.6 claude-sonnet-4-6 | 1M | $3 cache $0.3 | $15 | $1.17−61% cache $0.117 | $5.85−61% |
Claude Sonnet 5 claude-sonnet-5 | 1M | $2 cache $0.2 | $10 | $0.78−61% cache $0.078 | $3.9−61% |
Claude Haiku 4.5 claude-haiku-4-5 | 1M | $1 | $5 | $0.39−61% | $1.95−61% |
Claude Fable 5 claude-fable-5 | 1M | $10 cache $1 | $50 | $3.9−61% cache $0.39 | $19.5−61% |
Gemini 3.6 Flash gemini-3.6-flash | 1M | $0.75 cache $0.075 | $3.75 | $0.585−22% cache $0.0585 | $2.92−22% |
Gemini 3.5 Flash gemini-3.5-flash | 1M | $1.5 cache $0.15 | $9 | $1.17−22% cache $0.117 | $7.02−22% |
Gemini 3.5 Flash Lite gemini-3.5-flash-lite | 1M | $0.3 cache $0.03 | $2.5 | $0.234−22% cache $0.0234 | $1.95−22% |
Gemini 3.1 Pro gemini-3.1-pro-preview | 1M | $2 cache $0.2 | $12 | $1.56−22% cache $0.156 | $9.36−22% |
DeepSeek V4 Pro deepseek-v4-pro | 1M | $1.74 cache $0.0145 | $3.48 | $1.74 cache $0.0145 | $3.48 |
DeepSeek V4 Flash deepseek-v4-flash | 1M | $0.14 cache $0.0028 | $0.28 | $0.14 cache $0.003 | $0.28 |
DeepSeek V4 Flash (Free) deepseek-v4-flash-free | 1M | $0.14 cache $0.0028 | $0.28 | Free−100% cache Free | Free−100% |
Grok 4.6 grok-4.6 | — | $2 cache $0.5 | $6 | $2 cache $0.5 | $6 |
Prices are USD per 1M tokens. “list” is the official provider price; “Cradler” is your rate through Router.
Prices are USD per 1M tokens and track upstream rates. The live rate for any model is always what GET /v1/models returns for your key.
One line of code to switch
Cradler Router is protocol-compatible with the official SDKs — point the base URL at Router, keep everything else.
from openai import OpenAI
client = OpenAI(
base_url="https://router.cradler.ai/v1", # the only line you change
api_key="<your sk-cr-… key>",
)
response = client.chat.completions.create(
model="gpt-5.5",
messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)One command, every agent
Paste your key, copy the command, run it in a terminal. It writes the config for Claude Code, Codex, and Gemini CLI — plus OpenClaw, ZCode, and Cherry Studio when they're installed — then verifies your key against the Router. Built locally; the key never leaves this page.
npx @cradler/router-setup <your-router-key>Full protocol reference, agent guides, and billing endpoints in the API docs.
Built for production agent workloads
No silent downgrades
Every request reaches the exact model you asked for, over its native protocol — no substitution, no quality throttling. Verify any request's model, tokens, and cost via GET /v1/billing/requests/{id}.
Prompt caching preserved
Cached context passes through and bills at the cache-hit rate (about 10% of input price) — the savings that matter most for coding agents that resend a large context every turn.
Transparent balance billing
A prepaid USD balance, no subscription. Every top-up and every charge is itemized in your billing history, and your agents can query balance and per-request cost over the API.
Data stays private
Prompts and completions are used for routing, metering, and billing only — never for training, never sold.
Frequently asked questions
What is an LLM gateway?
An LLM gateway (or model router) is a single API endpoint that sits in front of many model providers: you authenticate once, call one base URL, and reach every model behind it. Cradler Router speaks the Anthropic, OpenAI, and Gemini protocols natively, so official SDKs and coding agents work unchanged — only the base URL and key change.
Which coding agents work with Cradler Router?
Claude Code, Codex, and Gemini CLI are configured by one command (npx @cradler/router-setup <key>); Cherry Studio imports with one click; OpenClaw and ZCode are auto-detected; Cursor, Trae, and WorkBuddy take the base URL and key in their model settings. Anything that speaks the OpenAI, Anthropic, or Gemini protocol works.
How much does the Claude API cost through Cradler Router?
Claude Sonnet 4.6 costs $1.17 per 1M input tokens and $5.85 per 1M output tokens through Cradler Router — about 61% below Anthropic's list price of $3/$15. Every Claude tier (Haiku, Sonnet, Opus) carries the same discount.
How much does the GPT API cost through Cradler Router?
GPT-5.5 costs $0.65 per 1M input tokens and $3.9 per 1M output tokens — about 87% below OpenAI's list price. The GPT-5 series is the deepest discount in the lineup.
Is there a free AI model API?
Yes — deepseek-v4-flash-free costs $0 and works with a zero-balance account, limited to 50 requests per account per day. It is the easiest way to test your integration before topping up.
How is Cradler Router priced?
Per token, in USD, per 1 million tokens — shown separately for input, output, and (where supported) cache hits. You top up a balance and usage draws it down. There is no subscription for model access.
Are these prices final?
Prices track upstream provider rates and refresh daily; every request settles at the live rate at the moment it runs. The authoritative rate for any model is always what the API returns — read GET /v1/models rather than hard-coding numbers.
How can it be cheaper than the providers?
Cradler Router aggregates supply across providers and routes each request to available capacity, so many models land well below first-party list prices — the GPT-5 series and Claude are the biggest savings. Some models (for example certain DeepSeek tiers and Grok) are already low-cost and are offered at parity. Retail never exceeds the official list price.
Do cache hits really cost that little?
Yes. For models that support prompt caching, repeated context is billed at the cache-hit rate rather than full input price. Coding agents like Claude Code reuse a large context every turn, so caching is where most of the real-world savings come from.
How do I pay?
Top up your account balance by card via Stripe, then create an API key — no separate provider account required. Alipay and WeChat Pay are coming soon.
Keep reading
Cradler Router overview
One key for Claude, GPT, Gemini and more — how the gateway works.
API reference
Base URL, auth, protocols, agent guides, and billing endpoints.
Claude Code setup
Two environment variables to run Claude Code on Router.
Cherry Studio one-click
Import every model into Cherry Studio with a single click.
Get a Router key
Create a Cradler account, top up by card, and every model on this page is one API key away.