Up to 96% below official API prices

Uncensored models. One API key.

All 26 models

Chat in the browser, or use one API key in Claude Code, Codex and your own apps. Prepaid credits, no subscription.

Uncensored
Or connect your tools

Per request 100K input + 20K output tokens

GPT-6.1 Sol $0.40$0.030−92% Claude Sonnet 5.5 $0.40$0.072−82% GLM 5.3 $0.067Abliterated DeepSeek V4.1 Flash 1017 $0.055Abliterated
Prices, savings and setup
Up to 96% below official API prices

How can I help you?

GPT, Claude, Gemini, DeepSeek and 22 more models in one chat and one API key.

Or connect your tools

Each message is paid from your credits at the model's per-token rate. By chatting you agree to our Terms and Privacy Policy.

One API key.

Pay up to 96% less than the official API

Same models, same APIs. Cost of one request (100K input + 20K output tokens):

Official list prices, checked 2026-11-10. Abliterated models are compared with their base model.

See all prices

Compatible with the tools you already use

One key, one base URL. OpenAI and Anthropic compatible.

FAQ

How much does RouterHub cost?

There is no subscription. You add prepaid credits and pay per token at each model's published rate, in USD per million input and output tokens. A typical chat message costs a fraction of a cent. See pricing.

Is there free credit?

Creating an account is free. Chat and the API are paid from prepaid credits. New accounts get $5 free on their first top-up of $20 or more. See pricing.

Which APIs does RouterHub support?

GPT and Codex models use the Responses API, Claude models use the Anthropic Messages API, and DeepSeek, GLM, Grok, Gemini, Qwen and Kimi use Chat Completions. One RouterHub key works for all of them.

Which coding tools work with RouterHub?

OpenCode, Continue, Cline, Claude Code and Codex. Codex needs a Responses model and Claude Code needs an Anthropic Messages model. See the setup guides.

What are uncensored and abliterated models?

Reduced-refusal variants of open models such as DeepSeek, GLM, Qwen and Kimi. Abliteration reduces refusals through weight-level changes. Browse uncensored and abliterated models.