OpenRouter presents itself as the unified interface for every model. The homepage says you generate anything through a single, unified interface, covering text, images, videos, and audio, and that there are no subscriptions. The pricing table then names Free, Standard, Business, and Enterprise, with platform fees of N/A, 5.5%, 8%, and fee discounts. Credits can be used with any model or provider after you create an API key.
Official website · openrouter.ai
OpenRouter fits an app that should call many providers from one key and can accept a platform fee on pay-as-you-go usage. The quickstart says one endpoint reaches hundreds of AI models through a single API endpoint, handles fallbacks automatically, and picks a cost-effective option. Auto-routing is No on Free and Yes from Standard upward. It is a weak fit if you need the FAQ answers on markup or training.
The homepage calls OpenRouter the unified interface for every model. It lists 500+ models, 80+ providers, and text, images, videos, and audio, and it says you generate anything through a single, unified interface. Signup is three steps: create an account, buy credits that can be used with any model or provider, then create an API key and start making requests. That key is fully OpenAI compatible. The quickstart is the call shape. OpenRouter gives you access to hundreds of AI models through a single API endpoint. You send standard HTTP requests to the /api/v1/chat/completions endpoint with a model slug and a messages array. The sample slug ~openai/gpt-sol-latest is an alias for the newest model in that family, and any slug can be substituted. Optional referer and title headers are for leaderboards. It includes.68 /M input tokens and $2.09 /M output tokens.
Official documentationThe homepage says you fall back to other providers when one goes down, and that OpenRouter runs at the edge for minimal latency between users and inference. The quickstart says it handles fallbacks automatically and picks the most cost-effective option for each request. Those two sentences do not name a plan. The pricing table is stricter. Auto-routing is No on Free and Yes on Standard, Business, and Enterprise. Budgets, prompt caching, and data-policy routing follow that split. EU in-region routing and US in-region routing are No on Free and Standard and Yes on Business and Enterprise. Read a Free experiment as the 50 requests/day row, not as the auto-routing row.
Official documentationThe homepage says better prices, better uptime, and no subscriptions. It includes the fee platform fees of Free N/A, Standard 5.5%, Business 8%, and Enterprise fee discounts. Standard and Business are pay-as-you-go. Those two columns list credit card, crypto, and more. Enterprise lists invoicing options. Free shows no payment options. Bring-your-own-key is a separate cell. Free is No. Standard and Business each say $25,000 of list price inference / month with no fees, then a 5% fee after. Enterprise says Custom. Free lists 25+ free models and 4 free providers, while the other columns list 500+ models and 80+ providers. Chat and API access are Yes on every column. The markup question has no answer, so these platform fees are not restated as a per-token markup.
Official documentationWorkspace limit is 5 on Free, 5 on Standard, 1,000 on Business, and Custom on Enterprise. Management API keys and admin controls are No on Free and Yes from Standard upward. Advanced notifications and workload identity federation are No on Free and Standard and Yes on Business and Enterprise. Support is community support, email support, or a dedicated support team in that order. SSO/SAML, contractual SLAs, and managed policy enforcement are Yes only on Enterprise. The DPA cell is via terms of service on Free and Standard, a signed copy on request for Business, and customizations for Enterprise. Rate limits are 50 requests/day on Free, passthrough from the providers on Standard and Business, and higher allowances on Enterprise.
Official documentationIs to create an API key and send one POST to the /api/v1/chat/completions endpoint, then change only the model slug. The quickstart says any slug can be substituted. Credits can be used with any model or provider, so the homepage treats one balance as spendable across the catalog. Before you raise volume, read that slug's own token price. It includes.68 /M input tokens and $2.09 /M output tokens.
A suggested first pass is the Free column only. It lists 25+ free models, 4 free providers, chat and API access, and 50 requests/day. Auto-routing, budgets, prompt caching, and data-policy routing are No. Do not expect the quickstart's cost-picking sentence to override a No cell. Use the free cap to learn the request shape. When you need fallbacks or a budget control, the table moves those to Standard, which is where the 5.5% platform fee is shown. It is not a load test.
Business also raises the workspace limit from 5 to 1,000. Its platform fee cell is 8%. If the requirement is SSO/SAML, a contractual SLA, or managed policy enforcement, those cells are Yes only on Enterprise, where the fee cell says discounts are available rather than a percent. The suggestion is how to read the columns.
Sign up. The homepage says you can set up an org for a team later.
Buy credits if you will leave the free cap. Credits can be used with any model or provider. Free is 50 requests/day.
Create an API key and start making requests. The homepage says the interface is fully OpenAI compatible.
POST to the /api/v1/chat/completions endpoint with a model slug and a user message. Check the plan column before you rely on routing.
The public platform fees, not a monthly subscription price. The homepage says no subscriptions. Standard and Business are pay-as-you-go. Enterprise says fee discounts are available. One catalog row was $0.68 /M input tokens and $2.09 /M output tokens.
Trying chat and the API inside a daily cap.
N/A.
Platform fee cell.
Routing and budgets without in-region locks.
5.5%.
Platform fee, pay-as-you-go.
Teams that need EU or US in-region routing.
8%.
Platform fee, pay-as-you-go.
Organizations that need SSO, SLAs, or invoicing.
Not verified.
Fee discounts available.
Do not treat a review's minimum top-up as if it were. FAQ answers on markup and training were not stored. BYOK on Standard and Business is $25,000 of list price inference / month with no fees, then 5%. Free BYOK is No.
You want one provider's endpoint instead of this multi-provider table.
OpenRouter's public pages are a router across 80+ providers, a platform fee, and a shared credit balance. Choose Groq when that routing table is not the job.
You need a hub for model files and datasets, not only a chat call.
Hugging Face, in its own public pages, is where people host models, datasets, and Spaces. Choose it when publishing a repo is the work and OpenRouter's chat-completions fee table is not. No extra Hugging Face price is added here.
You want the image-tagged builder instead of this fee table.
OpenRouter's homepage already includes images, video, and audio inside the unified interface. Pick Replicate when you are not buying OpenRouter credits or reading this platform-fee row.
The fee row says Free N/A, Standard 5.5%, Business 8%, and Enterprise fee discounts available. Standard and Business are pay-as-you-go. The homepage also says there are no subscriptions. The table does not show a monthly seat price.
SourceFree lists 25+ free models, 4 free providers, chat and API access, activity logs, a workspace limit of 5, and 50 requests/day. Auto-routing, budgets, prompt caching, data-policy routing, in-region routing, management API keys, and BYOK are No. Support is community support.
SourceCreate an API key, then send standard HTTP requests to the /api/v1/chat/completions endpoint. The sample uses a bearer key, a model slug, and a user message. Optional referer and title headers affect leaderboards. The homepage says the interface is fully OpenAI compatible.
SourceThe quickstart says OpenRouter handles fallbacks automatically and picks the most cost-effective option for each request. The pricing table marks auto-routing No on Free and Yes on Standard, Business, and Enterprise. Use the table when the two sentences disagree.
SourceFree is No. Standard and Business each allow $25,000 of list price inference / month with no fees, then a 5% fee. Enterprise is Custom. That cell is separate from the 5.5% and 8% platform-fee row.
SourceA pricing heading asks whether OpenRouter trains on customer data. The same is true of the markup question. What is shown is data-policy routing: No on Free, Yes on Standard and above.
SourceOct 6, 2026 · NavHunt editorial
NavHunt did not run OpenRouter, did not buy credits, and did not send a chat-completions request. This profile restates the homepage, pricing table, models catalog sample, and quickstart reviewed on 6 October 2026. This description is based on OpenRouter's official materials reviewed on 2026-10-06, including OpenRouter homepage, OpenRouter pricing table, OpenRouter model catalog, OpenRouter quickstart.
Compare the free request cap, Standard 5.5% fee, Business 8% fee, and Enterprise discounts using the public table.