A managed marketplace resells model access. One account and one key reach models from many underlying providers, with the marketplace handling credentials, billing and failover on your behalf. You pay it rather than the model vendors, and it takes a percentage — either on the tokens themselves or when you top up a prepaid balance.
What this category is
This is the fastest route from nothing to a working call against almost any model, and for evaluation work it is hard to beat. One signup replaces a dozen vendor accounts, each with its own billing relationship, rate limit and approval process. When the model you want to try was released this morning, this is the category that has it.
The cost is control over the data path. Your prompts pass through a third party that is not the model vendor, which adds a party to every privacy review and a link to every failure chain. Retention terms, zero-data-retention options and which underlying provider actually served a given request are the questions worth asking here, and they are the fields most often left unpublished.
Why you cannot compare these prices with the others
The headline number is a percentage, which makes it look expensive next to a flat fee and cheap next to nothing. Both readings are wrong at the wrong volume. A percentage is trivial while your spend is small and becomes the dominant line as you scale, which is the inversion the pricing guide works through with real figures.
The pricing guide works
through all four charging mechanisms — token markupMarkup: A percentage the gateway adds on top of what the model actually costs. Some charge none at all and make money elsewhere.MarkupA percentage the gateway adds on top of what the model actually costs. Some charge none at all and make money elsewhere., credit feesCredit or top-up fee: A cut taken when you add money to a prepaid balance, typically around 5%. Easy to miss because it is not a markup on tokens — but you pay it on every dollar you load.Credit or top-up feeA cut taken when you add money to a prepaid balance, typically around 5%. Easy to miss because it is not a markup on tokens — but you pay it on every dollar you load., per-seat, and self-hostedSelf-hosted: You operate the software on your infrastructure and own its uptime, patching and upgrades. Model hosting, telemetry and control planes may still involve third parties; commercial editions can have licence fees.Self-hostedYou operate the software on your infrastructure and own its uptime, patching and upgrades. Model hosting, telemetry and control planes may still involve third parties; commercial editions can have licence fees. — with figures computed at four real
workloads, and the cost estimator runs the same
model against your own volumes.
When this is the wrong category
If your data cannot leave your network, or a security review will ask who else touched the request, start with the open-source category instead. No marketplace answers that objection.
European multi-modality AI gateway: LLM chat behind an OpenAI-compatible endpoint plus OCR, speech, translation and vision APIs, billed at provider cost with a 5.5% fee on credit purchases.
France company · EU region available
Cost above the model bill
5.5% to top up
No markup on tokens, fee applies when you add funds
Models
1,158
across 33–68 providers
Failover
Spend limits
Logs
Caching
Guardrails
Best for Teams that need OCR, speech, translation or document parsing alongside LLM calls under one contract, one prepaid balance and EU data residency, and who are happy to build routing policy per request.
Zero-markup routing layer that fans a single Hugging Face token out to 18 third-party inference partners.
US company
Cost above the model bill
No markup
You pay the model provider’s own prices
Models
127
across 14–18 providers
Failover
Spend limits
Logs
Best for Teams calling open-weight models who want one token, one bill and genuinely zero markup across 18 serverless providers, with automatic failover they do not have to configure.
Hosted access to text, image and video models through one account.
Cost above the model bill
See pricing
Not published in a directly comparable form
Models
94
across 20+ providers
Failover
Logs
Best for Developers evaluating consolidated access to language and media models.
The other four categories
Products are sorted by what they actually are, not by what they are marketed as.
If none of the above is the shape of your problem, one of these probably is.
Do LLM marketplaces cost more than going direct to the model provider?
Almost always yes, in per-token terms, because the fee is added to a price you could have paid directly. Whether that is worth it depends on how many provider relationships it saves you and how much your time is worth. At low volume the fee is usually smaller than the cost of managing the accounts it replaces.
Is a marketplace fee a token markup or a top-up fee?
These are different mechanisms and the distinction matters. A token markup is a percentage added to every call. A top-up or credit fee is taken once when you add funds to a prepaid balance, so it does not compound with usage in the same way. Each product page records which applies.
Can I bring my own provider keys to a marketplace?
Some support it, which removes the resale margin and leaves you paying the model vendor directly for tokens. Support is uneven and sometimes carries its own fee, so the BYOK field on each product page states what was found and links to the page stating it.
Analytics, only if you say so
GatewayScore would like to record which comparisons, guides and tools people use, so the
research goes where it is actually read. No advertising, no data sold, no session recording,
and nothing is loaded until you choose. What gets recorded.