OpenRouter

What it is for

OpenRouter is a multi-provider chat + OpenAI-compatible API: the signed-out Chat composer (see the cover: Ask anything, model chips, New chat) is the same catalog you call over HTTP. Use it when the job is one key across many models, not when you want a single-vendor assistant.

It is not open source. Free exists (50 requests/day). Standard is pay-as-you-go credits with a 5.5% platform fee.

What makes it unique: the cover is the live openrouter.ai/chat composer — Flagship / coding / reasoning chips plus the prompt box — not a model-count poster.

Key features

• Chat composer — Ask anything, attach, model chips (cover).
• 500+ models / 80+ providers on paid plans (25+ free models on Free).
• OpenAI-shaped API — swap the base URL; fallbacks retry other providers.
• Credits — buy a balance; inference is billed at the provider list price.
• BYOK — Standard/Business include $25,000/month list-price inference with no OpenRouter fee, then 5%.

Who it's for

Best fit: developers who want one router and will read per-model token rows.

Poor fit: people who want a single branded chat seat (ChatGPT) or a local desktop runtime (LM Studio).

Pricing

Verified 2026-10-21 (Asia/Shanghai) from openrouter.ai/pricing. Inference is not marked up; the platform fee is charged when you buy credits. Re-check before you buy.

• Free — $0. 25+ free models; 50 requests/day; 5 workspaces; no card. Free models also 20 rpm.
• Standard — 5.5% fee on credit purchases. 500+ models; auto-routing; budgets; prompt caching. No minimum, no lock-in.
• Business — 8% fee on credit purchases. 1,000 workspaces; EU/US in-region routing; WIF.
• Enterprise — contract (SSO/SAML, SLAs, invoicing).
• Free-model daily cap — 50/day, rising to 1,000/day after the account has bought at least $10 of credits (FAQ).

Do not invent a $20/month chat seat. Tokens are billed at each model’s posted rate.

Limitations / costs

• Free is 50 requests/day on free models. Failed attempts count.
• The 5.5% / 8% fee is on credit buys, not a hidden per-token markup.
• Not a consumer memory chat. For a single-vendor assistant, see ChatGPT.
• Rows move. Pin a model ID if you need a stable price.

Alternatives

• Together AI — hosted open-model inference with printed $/1M rows.
• LiteLLM — self-host gateway; OSS $0.
• Poe — consumer multi-model catalog with App Store seats.

Official tutorial

Written path: Quickstart plus the signed-out Chat. We did not confirm an official OpenRouter-channel Getting started How-to YouTube id for embed.

Website

https://openrouter.ai

Source

Catalog row is curated (2026-10-21). Cover is the live OpenRouter Chat composer (Ask anything + model chips), not an OG model-count card. Dollars checked on openrouter.ai/pricing (Free 50/day; Standard 5.5% credit fee; $10 unlocks 1,000/day on free models). No official Getting started YouTube confirmed for embed.

Last verified

2026-10-21 (Asia/Shanghai) editorial check.