OpenRouter Review (2026)
A managed gateway to 500+ models with no subscription, charging 5.5% on top of provider prices. Two details to know - there are no volume discounts, and your credits expire after a year of non-use.
Rating
Starting Price
5.5% markup
Free Plan
Yes
SDKs & Frameworks
2
Deployment
3
Best For
Small teams and prototypes trying many models without wanting infrastructure, and anyone whose monthly provider spend stays below roughly $5,000 where the markup costs less than running a gateway.
Last Updated:
10 Things You Should Know About OpenRouter
- 1 Charges a 5.5% markup on provider prices with no subscription fee
- 2 The 5.5% credit fee applies to all purchases with no volume discounts
- 3 Credits expire after one year of non-use
- 4 BYOK mode reduces the fee to 0.5% but requires you to manage provider API keys
- 5 Provides access to over 500 models through a single OpenAI-compatible endpoint
- 6 27 models are available completely free, community-subsidised
- 7 There is no self-hosted option - it is a managed passthrough service
Pros & Cons
Pros
- ✓ Zero setup and zero infrastructure - one key, one endpoint, 500+ models
- ✓ No subscription, so cost scales purely with usage and there is nothing to pay when idle
- ✓ 500+ models is the broadest single-endpoint access available, including models you would otherwise need separate accounts for
- ✓ 27 free community-subsidised models make experimentation genuinely free
- ✓ BYOK at 0.5% is a sensible middle path once volume grows
- ✓ OpenAI-compatible, so migrating away is largely a base URL change
Cons
- ✕ 5.5% markup with no volume discounts, so the fee grows indefinitely with your spend and never improves
- ✕ Credits expire after one year of non-use, which is unusual and easy to forget
- ✕ No self-hosted option at all, so their availability is your availability
- ✕ Passthrough architecture means you are trusting a third party with every prompt
- ✕ No meaningful evaluation or observability, so it is one component rather than a stack
Features
The convenience is real
One API key. One OpenAI-compatible endpoint. 500+ models. No subscription.
For a team trying to work out which model suits a problem, that is worth more than the pricing page suggests. Comparing eight models normally means eight provider accounts, eight sets of credentials, eight billing relationships and eight client integrations. Here it is a string change.
27 models are available completely free, community-subsidised, which makes experimentation genuinely costless.
The economics turn against you, structurally
5.5% markup on provider prices, with no volume discounts.
That last clause is the important one. The fee never improves:
| Monthly spend | OpenRouter fee | What it buys you |
|---|---|---|
| $1,000 | $55 | Cheaper than running anything yourself |
| $10,000 | $550 | Roughly break-even against self-hosted infrastructure |
| $50,000 | $2,750/mo | Exactly the same service as at $1,000 |
A self-hosted LiteLLM deployment costs roughly $200 to $500 a month in infrastructure regardless of whether you push $1,000 or $50,000 through it. OpenRouter’s fee scales with your business indefinitely.
The crossover for most teams sits in the low thousands of monthly spend. Model it against your trajectory, not your current bill - the decision you want is the one that is still right in eighteen months.
BYOK mode drops the fee to 0.5% if you bring your own provider keys. That is a sensible middle path: you keep the unified endpoint and model catalogue, and take back credential management - which is much of the administrative burden a gateway existed to remove.
Credits expire after a year
Worth flagging on its own because it is unusual and easy to miss.
Credits expire after one year of non-use.
That catches teams who front-load credit purchases for convenience, or who run seasonal or project-based workloads that pause. Buy credits against near-term expected usage rather than in large blocks, and set a reminder if you have a balance sitting idle.
No self-hosted option
This is the structural risk. OpenRouter is a managed passthrough with no open-source build, so:
- Their availability is your availability. There is no fallback you control.
- Every prompt passes through a third party. For regulated workloads that may be disqualifying on its own.
The mitigation is partial but real: because the interface is OpenAI-compatible, an emergency migration is a base URL change rather than a rewrite. That limits the blast radius considerably - but it is a manual response during an incident, not automatic failover. Plan the fallback before you need it.
The intended path is through it, not to it
The most useful way to think about OpenRouter is as the first gateway rather than the last.
Both OpenRouter and LiteLLM expose OpenAI-compatible interfaces, so switching between them is typically a base URL and key change. That makes prototype on OpenRouter, move to self-hosted LiteLLM for production a genuinely realistic path rather than a theoretical one.
Which means there is very little to agonise over up front. Start on OpenRouter because it costs nothing to begin and removes all the setup. Move when the markup exceeds the cost of ownership. Neither step is expensive.
One caveat on the free models: community-subsidised is not a contractual guarantee. Which models are free, and on what terms, can change. Fine for prototyping; do not build a production dependency on a specific free model without a plan for it becoming paid.
Should you use it?
Use OpenRouter if you are prototyping, comparing models, or running under roughly $5,000 a month, and you do not want infrastructure.
Don’t use it if your spend is large enough that 5.5% exceeds the cost of self-hosting, you need a fallback you control, or prompts cannot leave your infrastructure.
Bottom line: the best on-ramp in this category and a poor long-term home for a high-volume application. Use it deliberately as a starting point, watch the crossover, and remember the credits expire.
Pricing model, fee structure, credit expiry and model coverage verified against vendor documentation and third-party comparisons on 3 August 2026. This is a researched directory entry - we have not yet instrumented this gateway with our reference application.
Pricing Plans
Standard
5.5% on provider prices
- No subscription fee
- Access to 500+ models
- No volume discounts
- Credits expire after one year of non-use
BYOK
0.5% fee
- Bring your own provider API keys
- Reduced fee, but you manage keys yourself
Free models
$0
- 27 models available free
- Community-subsidised
SDKs & Frameworks
Deployment
Eval Methods
Billing Unit
Our Verdict
OpenRouter is the fastest way to get access to a very large number of models, and for small teams it is close to unbeatable on convenience. One key, one OpenAI-compatible endpoint, 500+ models, no subscription, and 27 community-subsidised models that are genuinely free. If you are evaluating which model suits a problem, the ability to switch between hundreds by changing a string is worth real money in engineering time alone. The economics turn against you as you grow, and the reason is structural. The 5.5% markup carries no volume discounts, so the fee scales indefinitely with your spend while delivering nothing additional - at $50,000 a month that is $2,750 every month for the same service you got at $1,000. Two smaller details are worth knowing. Credits expire after one year of non-use, which is unusual enough to catch people out. And there is no self-hosted option, so their availability is your availability, with no fallback you control.
Similar Tools
Portkey
Teams that want one gateway to route, cache, budget and guardrail across many LLM providers, and are fine paying for the managed tier once they need real observability
TrueFoundry
Enterprises that need an AI gateway deployable on-premise or in their own cloud with governance features, and that want a managed product rather than operating open-source infrastructure.
Vercel AI Gateway
Teams already deploying on Vercel who want provider-agnostic model access at list price with no markup, and who do not need governance features that carry a per-request charge.
Kong AI Gateway
Organisations already standardised on Kong for API management, where bringing LLM traffic under existing governance matters more than LLM-specific features.
Frequently Asked Questions
When does the 5.5% stop being worth it?
Around the point where it exceeds the cost of running your own gateway, which for most teams is somewhere in the low thousands of dollars of monthly spend. At $1,000 a month the markup is $55, which is far less than the engineering time to run LiteLLM. At $10,000 it is $550, roughly comparable to production infrastructure plus some ownership overhead. At $50,000 it is $2,750 every month, and self-hosting is clearly cheaper. The important structural point is that there are no volume discounts, so the fee never improves - it grows with your business indefinitely while a self-hosted gateway's cost stays roughly flat. Model the crossover for your own trajectory rather than your current spend.
What is the credit expiry catch?
Credits expire after one year of non-use. That is unusual for a developer platform and easy to forget, particularly for teams that front-load credit purchases or run seasonal workloads. If you buy credits for a project that then pauses, you can lose them. The mitigation is straightforward - buy credits against near-term expected usage rather than in large blocks for convenience, and set a calendar reminder if you have a balance sitting idle. It is a small thing that costs real money when it bites.
What is BYOK mode and should I use it?
Bring your own keys - you supply your own provider API keys and OpenRouter's fee drops from 5.5% to 0.5%. It is a sensible middle path once your volume makes the full markup uncomfortable but you still want the unified endpoint and model catalogue. The trade is that you now manage provider accounts, keys and quotas yourself, which is much of the administrative burden that a gateway existed to remove. Think of it as a halfway house between full managed and self-hosting - you keep the routing convenience and give back the credential management.
What happens if OpenRouter goes down?
You go down, and there is no self-hosted fallback. This is the structural cost of a passthrough architecture with no open-source option - you depend entirely on their availability, and unlike a self-hosted LiteLLM deployment there is nothing you control that can keep serving. Because the interface is OpenAI-compatible, an emergency migration is a base URL change rather than a rewrite, which limits the blast radius. But that is a manual response during an incident, not automatic failover. If availability is critical, plan the fallback path before you need it.
Is the free tier really free?
Yes, with a caveat about how it is funded. 27 models are available completely free and are described as community-subsidised, which makes experimentation genuinely costless. That is a real benefit for prototyping and for evaluating whether a smaller open model handles your task. The caveat is that community-subsidised arrangements are not contractual guarantees - which models are free and on what terms can change, so do not build a production dependency on a specific free model without a plan for it becoming paid.
OpenRouter or LiteLLM?
OpenRouter to start, LiteLLM at scale, and the migration between them is deliberately easy. OpenRouter suits small teams trying many models without wanting infrastructure, with spend under roughly $5,000 a month. LiteLLM requires engineering ownership but takes no markup, so it wins decisively once your spend is large. Both expose OpenAI-compatible interfaces, so switching is often just a base URL and key change, which makes prototype on OpenRouter then move to self-hosted LiteLLM for production a genuinely realistic path rather than a theoretical one. There is no reason to agonise over this decision up front.