What is OpenKey's fee?

A flat 3% added to the provider's per-token price, on every paid model. Example: Claude Sonnet 4.5 costs $3.00 per 1M input tokens from the provider; through OpenKey you pay $3.09. The $0.090 difference is our entire margin.

Are there any fees besides the 3%?

No. No credit-purchase or top-up fee, no minimum spend, no monthly platform fee, no inactivity or expiring-credit fees, and no spread hidden in the token price. If a charge is not in the table on this page, it does not exist.

How do free models work?

25 of the 338 models are free: their providers charge nothing for them, and 3% of $0.00 is $0.00 — OpenKey adds no fee of its own. Free models are grouped at the top of the price table on this page.

Can I bring my own provider keys (BYOK)?

Yes. Provider keys you bring are encrypted at rest and used only to route your own requests. The security page documents how keys are stored, what gets logged, and what never does.

Are there volume discounts?

No. One rule for every model and every customer: provider price + 3%. No tiers, no negotiated rates, no exceptions — at any volume, the math printed on this page is the math you pay.

Pricing

LLM API pricing: the model's price plus a flat 3%.

No credit fees, no minimums, no spread hidden in the token price. All 338 models are priced below — every number includes the fee, with the provider price printed beside it.

Get API key All 338 prices ↓

Worked example

anthropic/claude-sonnet-4.5

Input · 1M tokens

$3.00 + 3% $0.090$3.09

Output · 1M tokens

$15.00 + 3% $0.450$15.45

FEE — FLAT, EVERY MODEL3%

HIDDEN CHARGES$0.00

3%
Flat fee: $0.00
Hidden fees: 338
Models: 25
Free models

The fee, compared

How 3% stacks up.

The honest version: self-hosting is cheaper if your time is free, and OpenRouter's documented fee is higher. Here is the same $100 of provider-price usage through each.

OpenKey

Fee on usage: 3% flat
Credit / top-up fee: $0.00
You operate it: No — managed
$100 of provider usage: $103.00

OpenRouter

Documented fee: 5% + $0.35
Applies: on credit purchases
You operate it: No — managed
$100 of provider usage: ≈ $105.35

LiteLLM (self-hosted)

Fee on usage: 0%
Credit / top-up fee: None — pay providers
You operate it: Yes — deploy, monitor, patch
$100 of provider usage: $100.00 + ops time

OpenRouter figure as documented by costbench.com: “5% markup plus $0.35 on top of provider costs” on credit purchases. LiteLLM's gateway is open source with no markup — you pay providers directly and run deployment, scaling, monitoring, and patching yourself, with no managed SLA. Terms change; verify with each vendor. Compare per-model prices in our live price table.

Every price

All 338 models, provider price beside ours.

Per 1M tokens. “You pay” is the provider price plus the flat 3% — check the arithmetic on any row. For sorting, search, and workload costs, use the price comparison tool.

Input /1MOutput /1M

ModelProviderYou payProviderYou payContext

Free models · 25

Paid models · sorted by input price

Variable pricing · routed per request

Provider prices sourced live from the catalog. Each row links to the model's page with full specs, benchmarks, and code samples — or browse by lab in the model explorer.

Questions

What is OpenKey's fee?: A flat 3% added to the provider's per-token price, on every paid model. Example: Claude Sonnet 4.5 costs $3.00 per 1M input tokens from the provider; through OpenKey you pay $3.09. The $0.090 difference is our entire margin.
Are there any fees besides the 3%?: No. No credit-purchase or top-up fee, no minimum spend, no monthly platform fee, no inactivity or expiring-credit fees, and no spread hidden in the token price. If a charge is not in the table on this page, it does not exist.
How do free models work?: 25 of the 338 models are free: their providers charge nothing for them, and 3% of $0.00 is $0.00 — OpenKey adds no fee of its own. Free models are grouped at the top of the price table on this page.
Can I bring my own provider keys (BYOK)?: Yes. Provider keys you bring are encrypted at rest and used only to route your own requests. The security page documents how keys are stored, what gets logged, and what never does.
Are there volume discounts?: No. One rule for every model and every customer: provider price + 3%. No tiers, no negotiated rates, no exceptions — at any volume, the math printed on this page is the math you pay.

More detail: how keys are stored · estimate your token costs · browse all models

One fee. Printed math.

338 models behind one OpenAI-compatible key, each at its provider price plus 3%.

Get API key Browse models