OpenRouter API pricing: every OpenRouter model, per 1M tokens

Estimate OpenRouter costs from the selected route, supported charge categories and routing constraints. Separate model usage from account-credit operations and any BYOK arrangement.

How OpenRouter pricing works

The model schema records token, request, media and cache categories, with conditional pricing overrides when an endpoint uses them. Official documentation.

Begin with the complete operation your application performs. A plain text completion, a search-enabled response and a media-generation task may involve different billable work. Preserve the unit beside every amount. Converting all rows into an undifferentiated token figure would hide the condition that makes the estimate meaningful.

The default catalog price is a starting reference, not a guarantee that every possible endpoint configuration produces the same amount. Keep routing choices in the scenario and inspect the eligible endpoint evidence when a narrow price difference determines the decision. A maximum-price condition can reduce eligibility as well as constrain cost.

BYOK connects your provider credentials through OpenRouter and has its own documented fee and fallback conditions. Official documentation.

Keep direct-provider spending and OpenRouter-account spending visible as separate records when using that mode. A familiar model charge appearing in two dashboards needs reconciliation, not an assumption that one is a duplicate. Document which account pays for inference and which settings permit a fallback to another billing route.

Full price list

Inspect the identifier, category and condition before comparing adjacent rows. Unknown data should remain an unresolved fact, and a non-token category should not be dropped merely because the application’s first calculator view is token-focused.

Estimate your cost

Interactive tool

Estimate your API costs

Your text and estimates stay in this browser. No API requests are sent to model providers.

Loading verified model records…

Preserve the workload in the OpenRouter cost calculator and record the routing assumptions beside the result.

Use the same input and output requirements when comparing routes. A response that needs a repair call or another model should be counted as a multi-call task. Otherwise the least expensive individual call can appear to win a comparison while producing a larger cost per accepted result.

Three worked scenarios

For a chatbot, decide whether continuity requires a fixed model or permits a fallback. Test the same conversation state against every allowed candidate. Include the cost of a follow-up model call if an initial answer fails validation rather than presenting the nominal first call as the whole interaction.

Interactive tool

Estimate your API costs

Your text and estimates stay in this browser. No API requests are sent to model providers.

Loading verified model records…

For summarization, identify whether requests repeat a stable document prefix or each contain unrelated material. Cache assumptions should follow that structure. Keep an uncached scenario available and inspect whether the actual serving route supports the cache category used in the estimate.

Interactive tool

Estimate your API costs

Your text and estimates stay in this browser. No API requests are sent to model providers.

Loading verified model records…

For a coding assistant, include the repository context, output patch and the checks required to accept it. A provider-routing change can affect operational behavior even when the model family remains familiar, so record the resolved route alongside usage and test results.

Interactive tool

Estimate your API costs

Your text and estimates stay in this browser. No API requests are sent to model providers.

Loading verified model records…

How to reduce cost on OpenRouter

Provider routing exposes price-oriented selection and maximum-price conditions, while prompt-caching behavior depends on the selected model and provider path. Official documentation.

Use a price constraint after specifying mandatory capabilities. A cheap endpoint that ignores a field essential to your output contract is not a useful optimization. Keep required parameter support explicit and test both ordinary and exceptional outputs before widening the routing set.

The caching guide describes provider-specific cache behavior and the usage categories that indicate reuse. Official documentation.

Measure reuse on a stable fixture and preserve the serving path while investigating differences. Avoid assuming a cache warmed through one route will produce the same result through every other endpoint. First establish the request structure and observed behavior; then use that evidence in the workload scenario.

Compare smaller candidates against a fixed acceptance criterion. If a simpler task can be routed to a less costly model, make that task boundary explicit in application logic and test the classifier or routing rule. A cost-saving policy should not quietly send difficult requests to a candidate that has never passed the relevant cases.

Billing, credits and limits notes

OpenRouter distinguishes account balance and per-key credit limits from request-rate restrictions. Official documentation.

Inspect the account setting that actually rejected a request before changing funds or traffic. A key cap, workspace budget and account balance can require different owner actions. Keep a record of which boundary was intended for the application so a quick operational fix does not silently remove its spending control.

Review free-route conditions and capacity and credit diagnostics. Do not let a fallback policy turn a deliberately free experiment into an unreviewed paid workload.

Price change history

Recalculate the previous request assumptions when a route price changes. Keep the rate effect separate from a different resolved model, a changed cache share or a longer generated response.

Last verified 12 Sep 2026 · Source ↗

Frequently asked questions

Sources

• OpenRouter quickstart ↗
• Models and pricing schema ↗
• Authentication ↗
• Credit and rate limits ↗
• Errors and debugging ↗
• Provider routing ↗
• Free model variants ↗
• Free router ↗
• Workspace budgets ↗
• Prompt caching ↗
• BYOK ↗
• Provider logging ↗
• Batch quickstart ↗
• Latest aliases ↗

Last verified 12 Sep 2026 · Source ↗