What it is
OpenRouter is a single API for 500+ AI models from 80+ providers, including OpenAI, Anthropic, Google and open-weight model hosts, with one key and one balance. It uses the same request format as the OpenAI API, so most existing code works by changing the base URL, and it can fall back to another provider when one is down.
Model calls cost the provider's list price with no markup. OpenRouter charges a fee when you buy credits: 5.5% on Standard, 8% on Business, and negotiated rates on Enterprise. Bring-your-own-key usage is free up to $25,000 of list-price inference a month, then 5%. Free models are capped at 20 requests a minute and 50 a day, rising to 1,000 a day once you have bought $10 of credits.
Prompts and completions are not stored unless you opt in, though request metadata is. Zero Data Retention routing is available on every plan, EU and US in-region routing on Business and Enterprise, and SSO and SCIM on Enterprise. OpenRouter lists SOC 2 Type II on its trust centre.
Our review
OpenRouter turns a dozen provider accounts into one key, one balance and one OpenAI-compatible endpoint, and for most teams that is worth the fee. Its pricing page lists 500+ models from 80+ providers, and when one provider for a model is down or slow it routes the request to another. The value is highest early in a project, before you know which model you will settle on.
The fee is honest but easy to underestimate. Inference is billed at provider list price, and OpenRouter says plainly that it does not discount inference; the 5.5% (Standard) or 8% (Business) is charged on credit purchases. A team spending $40,000 a month pays about $2,200 in fees on Standard, or $3,200 on Business. Bring-your-own-key changes the maths: the first $25,000 of list-price usage a month carries no fee and the rest 5%, so teams with their own provider contracts keep their negotiated rates and use OpenRouter mainly for routing and fallback.
Data handling is better than the "another middleman" worry suggests. Prompts and outputs are not stored unless you opt in, though token counts and latency are, and OpenRouter samples a small number of prompts for categorisation, stored anonymously if you have not opted in. Zero Data Retention can be enforced per model family or per request with provider.zdr, but the switch has side effects: enforcing it for Anthropic models removes Anthropic's first-party endpoints and leaves Bedrock and Vertex, which narrows your fallback options.
Latency is the hidden cost. OpenRouter publishes no millisecond figure for its overhead, and its own docs say cold edge caches in a new region, and a credit balance in single-digit dollars, both add latency. At large, steady volume on one model, going direct to the lab or to a dedicated host such as Together AI or Fireworks AI removes the fee and the extra hop.
Where it earns its place
A two-engineer team building a chat app with a model picker
Offer Claude, GPT and open-weight models without five vendor contracts
One key and the OpenAI SDK with a changed base URL; fallback covers provider outages. On Standard, $5,000 a month of usage costs about $275 in credit fees, less than the engineering time to maintain separate integrations.
An ML engineer choosing a model for a document-extraction pipeline
Run the same evaluation set across a dozen models before committing to one
Swaps the model string per run and compares cost and quality on one bill. Once a winner is chosen, moving that workload to the provider directly, or onto bring-your-own-key, removes most of the fee.
A procurement lead at a 200-person EU fintech
Approve an LLM gateway that passes a data-protection review
Business (8% fee) adds EU in-region routing through eu.openrouter.ai; SSO, SCIM, contractual SLAs and a signed DPA need Enterprise. ZDR is enforced in account settings, and the SOC 2 Type II report is requested through the trust portal.
Before you commit
- Setup
- Create a key, buy credits and point an existing OpenAI SDK at OpenRouter's base URL; there are no per-provider sign-ups.
- Integrations
- OpenAI-compatible API, so most SDKs and frameworks that speak OpenAI work unchanged; bring your own provider keys if you have them.
- Data and privacy
- No prompt or completion storage unless you opt in; metadata is kept. ZDR routing on all plans, EU and US in-region routing on Business and Enterprise, SOC 2 Type II. The signed DPA applies to Enterprise accounts only.
- Cost at scale
- At $40,000 a month of list-price inference, Standard adds about $2,200 and Business about $3,200 in fees. With your own keys at the same volume, the fee is 5% of the $15,000 above the free allowance, about $750.
- Limits
- Paid models inherit each provider's rate limits; free models are capped at 20 a minute and 50 a day (1,000 after a $10 purchase), and failed attempts count. No latency overhead figure is published.
What's good
- 500+ models behind one OpenAI-compatible API
- List-price inference; the fee applies only to credit purchases
- Provider fallback and ZDR routing on every plan
- Bring-your-own-key free up to $25,000 a month
What's not
- 5.5% to 8% credit fee adds up at volume
- No published latency overhead; low balances slow requests
- EU routing needs Business; SSO and SLAs need Enterprise
- Free models capped at 50 requests a day until you spend $10
Best for
Developers and small teams who want to test, compare or offer many AI models through a single integration.
Alternatives
OpenRouter: common questions
- Is OpenRouter worth it?
- OpenRouter is the sensible first integration for any app that needs more than one model; at steady high volume on a single model, its credit fee becomes the thing to engineer away. We rate it 4.6 out of 5.
- How much does OpenRouter cost?
- Pay provider list prices for models. Fee on credit purchases: 5.5% (Standard) or 8% (Business). Bring-your-own-key free for the first $25,000/month of usage, then 5%. Free models: 50 requests/day, or 1,000/day after a $10 purchase.
- What are the downsides of OpenRouter?
- 5.5% to 8% credit fee adds up at volume; no published latency overhead; low balances slow requests; eU routing needs Business; SSO and SLAs need Enterprise; free models capped at 50 requests a day until you spend $10.
- Who is OpenRouter best for?
- Developers and small teams who want to test, compare or offer many AI models through a single integration.
- How much does OpenRouter cost for a team or in production?
- At $40,000 a month of list-price inference, Standard adds about $2,200 and Business about $3,200 in fees. With your own keys at the same volume, the fee is 5% of the $15,000 above the free allowance, about $750.
- Is OpenRouter safe for company data?
- No prompt or completion storage unless you opt in; metadata is kept. ZDR routing on all plans, EU and US in-region routing on Business and Enterprise, SOC 2 Type II. The signed DPA applies to Enterprise accounts only.
- What are the best OpenRouter alternatives?
- The closest alternatives we have reviewed are Together AI, Fireworks AI, Groq and Hugging Face.
Answers checked Oct 22, 2026.
Make OpenRouter? Show this rating on your site
