# Liquid Inference compared with routers and gateways

Liquid Inference is an exchange. Providers post offers for each model at their own prices, and each request starts an auction among the offers that meet your rules. Most other services are routers or gateways. A router sells each model at a list price. A gateway sends your request to a provider account that you hold.

These pages compare the exchange with eight of these services. The facts about each service come from its own pages, checked on 10 November 2026. Each page links its sources.

## Summary

| Service                                                 | Price                       | Fee                           | Cap for each job | Preview                         | Open to sellers         |
| ------------------------------------------------------- | --------------------------- | ----------------------------- | ---------------- | ------------------------------- | ----------------------- |
| **Liquid Inference**                                    | Each provider               | 5%, or a floor for each token | Yes              | Yes, with price caps            | Yes, at their own price |
| [OpenRouter](https://liquidinference.ai/compare/openrouter.md)                       | List price                  | 5.5% on credit                | No               | No                              | Yes, by application     |
| [Requesty](https://liquidinference.ai/compare/requesty.md)                           | List price                  | 5% on credit                  | No               | No                              | No                      |
| [RouterPlus](https://liquidinference.ai/compare/routerplus.md)                       | List price, or the provider | None                          | No               | Partly, the order with no price | Yes, by application     |
| [Vercel AI Gateway](https://liquidinference.ai/compare/vercel-ai-gateway.md)         | List price                  | None, plus card fees          | No               | No                              | No                      |
| [Cloudflare AI Gateway](https://liquidinference.ai/compare/cloudflare-ai-gateway.md) | List price                  | 5% on credit                  | No               | No                              | No                      |
| [Portkey](https://liquidinference.ai/compare/portkey.md)                             | Your contracts              | A plan, from $0               | No               | No                              | No                      |
| [LiteLLM](https://liquidinference.ai/compare/litellm.md)                             | Your contracts              | None                          | No               | Partly, from typed counts       | No                      |
| [Helicone](https://liquidinference.ai/compare/helicone.md)                           | List price                  | A plan, from $0               | No               | Partly, from typed counts       | No                      |

## What the exchange does differently

- **Providers compete for each request.** Each provider quotes its own price, and the auction buys the lowest cost. The cost is the estimated charge plus the value of the predicted wait. See [Price and wait](https://liquidinference.ai/docs/price-and-wait.md).
- **A hard cap for each request.** `exchange.cap_usd` is the most that one request can cost. Settlement never exceeds it, and the provider carries any usage above it. See [Purchasing rules](https://liquidinference.ai/docs/rules.md).
- **A dry run before you buy.** `POST /v1/exchange/dry-run` shows the estimated charge and the price cap of each offer, and why the other offers are out. It places no job and holds no funds.
- **A receipt for each job.** It shows the provider, the offer rates, the fee and the charge, so you can check each charge. See [Receipts](https://liquidinference.ai/docs/receipts.md).
- **No charge for a failure.** A job that fails before any output pays nothing. If a provider fails after the first output, you get a full credit.

## Where another service fits better

- **You hold contracts with your providers.** A gateway such as Portkey, LiteLLM or Cloudflare AI Gateway puts controls in front of your own accounts.
- **You must run the software yourself.** Portkey, LiteLLM and Helicone are open source. The exchange is hosted only.
- **You want no fee on tokens.** RouterPlus and Vercel AI Gateway sell at list price with no fee on tokens. On the exchange, the price can be below the list price, and the fee is 5%. See [Fees](https://liquidinference.ai/docs/fees.md).
- **You send most traffic through your own cloud keys.** OpenRouter charges no fee on the first $25,000 a month, and Requesty and RouterPlus charge none.

## Move to the exchange

Each service takes the OpenAI format, so a move changes the base URL, the key and the model names. The guides in [Move from…](https://liquidinference.ai/docs/move-from-openrouter.md) have each step for OpenRouter, LiteLLM and Portkey.