Every model behind one key.
Very different bills.

31 products can sit between your app and the models — LLM gateways, model routers, and the cloud platforms that act like both — and their headline pricing hides the part that actually costs you money: top-up 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., per-seat charges, retention surcharges, and markupMarkup: A percentage the gateway adds on top of what the model actually costs. Some charge none at all and make money elsewhere. on tokensToken: Roughly three quarters of a word. Billing is per million tokens, and output tokens usually cost several times more than input.. This is a side-by-side LLM gateway comparison covering every one of them, with each figure linked to the vendor's own page.

Products tracked
31
Charge no markup
17
Can be self-hosted
17
Figures linked to a source
2125

The choice is not really about models

Almost every product here can call almost every model. What separates them is who runs the software, what they add to the price, and how much control you get over the data path. Sort out those three and the shortlist writes itself.

01

Who operates it

A managed service is running before lunch and costs a percentage forever. A 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. gatewayGateway: A single endpoint you send all your AI requests to, which then forwards them to whichever model you asked for. One integration instead of one per vendor. has no vendor fee and costs you an on-call rotation. There is no third answer, and picking wrong is the most expensive mistake on this page.

Work out which side you are on →
02

What it adds to your bill

"Zero markupMarkup: A percentage the gateway adds on top of what the model actually costs. Some charge none at all and make money elsewhere." and "no fees" mean different things at different vendors. One charges 5.5% when you add funds. One is free until you want your data deleted. One is free per-request but $20 per developer per month. At scale these are not rounding errors.

Model your own numbers →
03

What your auditors need

A SOC 2SOC 2: An audit of a company’s security controls. Type 1 checks the design at a point in time; Type 2 checks they actually operated over months. Procurement teams ask for this constantly. report, a signed HIPAAHIPAA BAA: A signed contract required before you can put US health data through a vendor. Without one you cannot legally use the service for protected health information, regardless of how secure it is. agreement, EU-only data residencyData residency: A guarantee about which countries your data is processed in. Usually matters for EU obligations or public-sector contracts., or a guarantee that nothing is retainedZDR — zero data retention: A commitment that your prompts and responses are not stored after the request completes. Frequently a paid add-on or an enterprise-only option rather than the default.. These are pass-or-fail requirements, and they eliminate options far more decisively than any feature comparison.

Filter by requirement →

Five kinds of product, often confused

These get compared as if they were interchangeable. They are not. Knowing which category you need eliminates most of the catalog immediately.

Managed marketplace

5

One account and one key gets you hundreds of models from dozens of providers. Fastest way to start, widest catalog, least control over the data path.

See all 5 →

Managed gateway

9

A hosted control plane in front of the models: routing, caching, guardrails, spend limits, and logs. You trade a little simplicity for governance.

See all 9 →

Open source

9

You run the gateway yourself. Full control over the data path and no per-request vendor fee, in exchange for operating it.

See all 9 →

Cloud platform

5

A hyperscaler surface offering several vendors models under one contract, one bill, and one identity system. Strong compliance story, limited to that cloud.

See all 5 →

Inference provider

3

Hosts open-weight models on its own hardware. Often the cheapest or fastest route to a specific open model, but it is one source, not a router.

See all 3 →

Where to start, one per category

A deliberate selection rather than a ranking: one product from each category, all available to customers in the United States, chosen to show how differently the five kinds are priced. How these were picked →

All 31 products
Managed marketplace Managed only

Hosted marketplace that routes one OpenAI-compatible API to models from many inference providers.

US company · EU region available

Acquisition pending

Cost above the model bill
5.5% to top up
No markup on tokens, fee applies when you add funds
Models
458
across ~83 providers
  • Failover
  • Spend limits
  • Logs
  • Caching

Best for Teams that want the broadest possible model and provider catalog behind one OpenAI-compatible key with unified billing.

Managed gateway Managed only

Vercel-operated gateway that routes AI SDK and OpenAI-format requests to many providers with zero token markup.

US company

Cost above the model bill
No markup
You pay the model provider’s own prices
Models
413
count not published
  • Failover
  • Spend limits
  • Logs
  • Caching

Best for Teams already shipping on Vercel with the AI SDK who want zero token markup and no credit-purchase fee.

Open source Self-host only MIT Security incident on record

Open-source AI gateway and Python SDK that puts one OpenAI-compatible API in front of many LLM providers.

US company

Cost above the model bill
Self-hosted
OSS is free (MIT); you pay only for your own containers plus Postgres and Redis. Enterprise features (SSO, RBAC, JWT auth, SCIM, audit logs, support SLAs) require a paid LiteLLM commercial license whose price is not published.
Models
—
count not published
  • Failover
  • Spend limits
  • Logs
  • Semantic cache
  • Guardrails

Best for Platform teams that want a free, self-hosted, maximally broad provider abstraction with virtual keys and budgets, at moderate request volumes.

Cloud platform Managed only

AWS-managed service for calling foundation models from 19 model providers through one AWS API.

US company

Cost above the model bill
See pricing
Not published in a directly comparable form
Models
~100
count not published
  • Caching
  • Guardrails

Best for Teams already standardized on AWS that need many model vendors behind one IAM-governed, compliance-attested API.

Inference provider Managed only

Inference provider running open-weight models on its own LPU hardware for very high output speed.

US company · no EU region

Cost above the model bill
See pricing
Not published in a directly comparable form
Models
11
count not published

Best for Latency-sensitive, user-facing applications on open-weight models where output speed matters more than catalog breadth.

Recently changed

Site releases, published model and pricing updates, and sourced provider edits — newest first.

Full history

A few questions, one shortlist

The wizard rules out anything that cannot meet your hard requirements, then ranks what is left against what you said matters. It shows you why each one placed where it did, and what is wrong with your top pick.