22 October 2026 · Site Methodology & Corrections · 5 min read
Our OpenRouter pricing page has said "credit top-up fee ~5%" since I first wrote it in July. I went back today to re-check it against the live openrouter.ai/pricing page while cleaning up the AI-API section, and that number is gone. In its place: two explicit tiers, 5.5% on Standard and 8% on Business, plus "fee discounts available" for Enterprise through sales. Nobody rounds 5.5% down to "~5%" on purpose -- it just means the page was written when the fee actually was closer to flat, and the tiers got added later without anyone (including me) coming back to check.
Token usage itself is untouched -- OpenRouter still passes through the upstream provider's price with $0 markup, which is the whole reason to use it over building your own router. The fee only hits once, when you buy credits:
| Étage | Frais de recharge de crédit | Payment methods |
|---|---|---|
| Standard (pay-as-you-go) | 5.5% | Card, debit, AliPay, crypto (USDC) |
| Business (pay-as-you-go) | 8% | Same as Standard |
| Entreprise | Negotiated discount | Invoicing / purchase order via sales |
No minimum spend on either pay-as-you-go tier, and no separate crypto discount -- USDC sits in the same fee bracket as a card, at least on the page as published today. Which bracket an account lands in isn't spelled out as a hard dollar threshold on the pricing page itself; it's a function of account settings under Settings → Credits, so check there rather than assuming Standard by default if you're running meaningful volume.
A flat "~5%" and a real 5.5%/8% split sound like a rounding error until you put a number behind it. To end up with $1,000 of usable credit:
| Étage | Fee on $1,000 usable | You actually pay |
|---|---|---|
| Old number (~5%) | ~$50 | ~$1,050 |
| Standard (real: 5.5%) | ~$55 | ~$1,058 |
| Business (real: 8%) | ~$80 | ~$1,087 |
At $10,000/month in LLM spend running through OpenRouter, that's $550/month in fees on Standard versus $800/month on Business -- a $250/month gap that has nothing to do with which models you call, only which tier your account sits in. Run your own model mix through the Calculateur de coût de l'API IA first, then add the top-up fee on top of whatever that number comes out to.
The fee is structurally invisible in the part of the product most people actually look at -- the per-model price list. Every model's $/1M-token rate is identical whether you're on Standard, Business or Enterprise, because that part really is pure pass-through. The fee only shows up on the Settings → Credits top-up screen, one click away from where anyone comparing "OpenRouter vs going direct" is actually looking. I wrote the ~5% number down once in July from that same screen and then never had a reason to open it again -- the same failure mode as the Groq/Cerebras model-list change last week: a number that was correct on the day I checked it, quietly aged out, and nothing forced a re-check until I went looking for an unrelated reason.
Updated the OpenRouter guide's pricing table to show both tiers instead of one flat number, added the Enterprise negotiated-discount row, refreshed the "last verified" date, and corrected the FAQ answer and the "when it's worth it" paragraph that were both still quoting ~5%. The core advice doesn't change: if you're running one model at serious volume, going direct to OpenAI, Anthropique ou Groq avoids the fee entirely. If you're switching between models or want fallbacks, the fee is the price of that flexibility -- just budget for 5.5-8%, not 5%.
Nouveau dans la tarification de l'API basée sur l'utilisation en général ? Commencez par le guides gratuits sur les coûts des API.
Déployez-le vous-même : DigitalOcean — 200 $ de crédit gratuit ↗ · VPS Hostinger ↗
Checked directly against openrouter.ai/pricing, 2026-10-22. Reference figures, not live quotes -- confirm current terms and your own account's tier directly with OpenRouter before budgeting.