What affects cost
- Selected model and provider endpoint
- Prompt, completion, cache, image, search, and reasoning units
- Credit purchase fees
- BYOK request volume after the free allowance
- Fallback attempts and routing behavior
LLM API
OpenRouter API pricing
OpenRouter combines many model providers behind one API while separating underlying inference rates from credit-purchase and BYOK platform fees.
Decision summary: Consider OpenRouter when unified model access, routing, fallbacks, and centralized usage records justify an additional billing layer.
Pricing overview
Underlying model inference is billed at the listed provider rate; platform fees depend on credit purchase, BYOK usage, and account plan.
| Plan or model | Pricing posture | Decision note |
|---|---|---|
| Pay-as-you-go inference | Model-specific listed price; no inference markup | OpenRouter states that underlying provider inference prices are passed through. Prompt, completion, request, image, search, reasoning, and cache charges can vary by model. |
| Credit purchase fee | 5.5% fee; $0.80 minimum | The fee applies when purchasing credits and is separate from the model inference rate. |
| BYOK | First 1M BYOK requests/month free; 5% after | After the free monthly request allowance, OpenRouter documents a 5% fee based on what the same model and provider would cost on OpenRouter. |
| Free model access | Free models with account-dependent rate limits | The official pricing page lists 50 requests/day for the Free plan; documentation notes higher free-model limits after qualifying credit purchases. |
| Enterprise | Model rates plus negotiated platform terms | Enterprise pricing can include fee discounts, invoicing, support, policy controls, and optional dedicated limits. |
Lower-cost options from the same provider
- Use a lower-cost model or provider endpoint from the OpenRouter catalog after task testing
- Use free models only when their rate limits and availability fit the workload
- Pin routing and fallback rules when cost predictability matters
Alternative providers or products
- Direct OpenAI API access
- Direct Anthropic API access
- Direct Google Gemini API access
- Together AI or GroqCloud for supported open-model inference
Best for
- teams evaluating models across providers through one API
- applications that value fallback and routing options
- buyers that need centralized model usage records
Not ideal for
- teams that only use one direct provider and need no routing layer
- budgets that ignore credit-purchase or BYOK fees
- workloads without model and provider pinning requirements
FAQ
Questions teams ask before choosing
Does OpenRouter add a markup to model inference prices?
OpenRouter says underlying provider inference pricing is passed through without markup, while credit purchases and some BYOK usage have separate platform fees.
What should teams include in an OpenRouter budget?
Include model-specific token and feature charges, credit-purchase fees, BYOK request volume, routing behavior, fallbacks, and any enterprise terms.