GPT-5.6 · Claude Fable 5 · Kimi K3 — all live

One API key for 37 models—including Claude, GPT and Gemini.

Add $5 and pay vendor list price with no markup or top-up fee. Set hard per-key budgets so leaked keys and runaway agents cannot create surprise bills.

$5 minimum · no subscription · credit never expires

# same endpoint — swap the model string
curl https://api.routerplex.com/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERPLEX_KEY" \
  -d '{ "model": "gpt-5.6-terra" }'
live catalog
29 models
Anthropic
5 models
MiniMax
4 models
OpenAI
3 models
Alibaba
3 models
Chat
28 models
Images
1 model
Markup
0%

FIG 01 — Live model catalog by provider.
Every model behind the same endpoint, key, and bill.

37 models · 14 providers · two API formats

OpenAIOpenAIAnthropicAnthropicGoogleGoogleDeepSeekDeepSeekAlibabaAlibabaMoonshotMoonshotZhipuZhipuMiniMaxMiniMaxByteDanceByteDanceStepFunStepFunXiaomiXiaomi

§ 02 — Why RouterPlex

Stop juggling SDKs,
keys, and billing pages.

Every provider has its own SDK, pricing model, and quota system. RouterPlex collapses all of that into one account.

Without RouterPlex
  • 11 provider accounts to manage
  • 6 SDKs, 6 auth schemes
  • 11 separate invoices a month
  • quota limits tracked in a spreadsheet
  • one provider outage → your app is down
integration time: weeks, forever
With RouterPlex
  • 1 account, 1 API key
  • the OpenAI SDK you already use
  • 1 prepaid balance, per-token billing
  • hard spend limits on every key
  • automatic fallbacks between models
integration time: base_url = "https://api.routerplex.com/v1"

11 invoices become 1 · 6 SDKs become 0 · spend limits, scoped keys, and auto fallbacks on every account

§ 03 — Pricing

Pay per token.
No markup games.

Model
Provider
Context
Input / 1M
Output / 1M
Anthropic
claude-opus-4-8
Anthropic
1M
$5.00
$25.00
OpenAI
gpt-5.5
OpenAI
256K
$5.00
$30.00
Anthropic
claude-sonnet-4-6
Anthropic
1M
$3.00
$15.00
Google
gemini-3.1-pro
Google
1M
$2.00
$12.00
DeepSeek
deepseek-v4-pro
DeepSeek
1M
$0.43
$0.87
Moonshot
kimi-k3
Moonshot
1M
$3.00
$15.00
Alibaba
qwen3.7-max
Alibaba
1M
$2.50
$7.50
OpenAI
gpt-image-2image
OpenAI
$5.00
$30.00

featured 8 of 37 live models · prices in USD per 1M tokens, billed per token

§ 03b — The switch

Every $100 you route through us
buys $100 of tokens.

On OpenRouter, topping up $100 by card costs $105.50. Here it costs $100.

RouterPlexOpenRouter
Fee when you top up$0.005.5% by card ($0.80 min) · 5% by crypto
Per-token pricesVendor list priceVendor list price
Balance can go negativeNever — requests stop at $0Yes — 402 errors until you repay
Plans with bonus creditsUp to +25% ($125 for $100)Not offered
Hard per-key budgetsEvery keyEvery key

OpenRouter figures from their published docs, July 2026. Read the full comparison →

§ 04 — Plans

Pay per token.
No subscription required.

Top up when you want — a top-up is a one-time payment, not a plan. Subscribe only when the bonus credits beat topping up by hand.

Pay as you go — the default
Top up from $5, use every model at vendor list price.
60 RPM · 250K TPM per key · 3 keys · credit never expires
Start with $5

§ 05 — How it works

Three steps.
Under a minute.

No subscription, no monthly fee, no lock-in. Top up from $5 and pay per token.

Start now
  1. 01

    Create your account and top up

    Sign up with email or GitHub, then top up from $5 by card or crypto. Your balance is the only thing you ever pay.

    + $5.00 balance · one-time top-up
  2. 02

    Mint a scoped key

    Create a key per project. Attach a hard budget, a model allowlist, or rate limits — spend stops the instant a cap is hit.

    sk-rpx-•••••••• · budget $10 · all models
  3. 03

    Point your SDK at us

    Keep your OpenAI client. Change the base URL, pick any model from the catalog, ship.

    base_url = "https://api.routerplex.com/v1"

§ 06 — The deal

Pay for tokens.
Nothing else.

Every model is billed at the vendor's public list price, per token, from a prepaid balance you top up by card or crypto. 37 chat and image models across 14 providers — and the meter only runs when you call one.

RouterPlexstatement / monthly
Monthly fee$0.00
Per-seat charge$0.00
Minimum spend$0.00
Top-up fee$0.00
Markup over list price$0.00
Tokens you usedlist price
Fixed cost / month$0.00

* Credit never expires. When your balance runs out, requests stop — they never overdraft.

37
models live now
11
providers
0%
markup over list price
$0.00
fee on top-ups

§ 06b — Questions

Fair questions, straight answers.

What is RouterPlex?

RouterPlex is an AI gateway that gives you one API key to call 37 models from 14 providers — including Claude, GPT, and Gemini — through OpenAI- and Anthropic-compatible endpoints, with usage-based pricing and hard per-key spend limits.

How does RouterPlex pricing work?

You pay per token at the vendor's published list price with 0% markup and no fee on top-ups. Top up from $5 by card or crypto — no subscription required, and your balance never goes negative. Optional monthly plans (Plex Lite, Pro, Max) include bonus credits worth more than their price plus higher rate limits.

If the markup is 0%, how does RouterPlex make money?

Two ways: optional monthly plans that sell higher rate limits (their bonus credits are still worth more than the plan costs), and the float on prepaid balances. Token prices stay at vendor list price — we don't make money on your usage.

Is RouterPlex compatible with the OpenAI SDK?

Yes. Point OpenAI clients at api.routerplex.com/v1. RouterPlex also supports the Anthropic SDK and /v1/messages format at api.routerplex.com; both formats reach every chat model.

How is RouterPlex different from OpenRouter or LiteLLM?

RouterPlex is a hosted, ready-to-use gateway: one key, vendor list prices with 0% fee on top-ups (OpenRouter charges 5.5% by card), per-key budgets, and a prepaid balance that never goes negative — no proxy to self-host and no lock-in. Read the full OpenRouter comparison →

§ 07 — Get started

Start routing in
under a minute.

Pay $5 and receive $5 in usable credits. Add a hard budget to every key before your first request.

one key · every model · billed per token