AI API Cost Estimator
Estimate monthly text-model API spend from requests, input and output tokens, caching and retries. Edit USD prices and separate tool and infrastructure costs.
Calculate locally. No account or contact details required. Inputs are not saved; copy or print your report before leaving.
Plan your AI workload
Start with an example workload, choose a model, then tune usage. Prices are USD per million tokens; estimates update immediately.
Workloads are illustrative examples, not provider recommendations.
Search all options. Arrow keys move through matches; Enter selects and Escape closes.
Loading model prices from the public catalogue…
Original user operations before additional retries.
Include prompts, retrieved documents and context.
Include reasoning tokens when your provider bills them.
Advanced options — caching, retries and prices
Price assumptions
Choose Custom prices in the model selector to edit these rates. Selecting a live model restores its catalogue rates.
How this tool works
- Monthly attempts = monthly requests × (1 + additional retry percentage / 100).
- Input tokens are split into uncached and cached shares. Each token total is multiplied by its USD price per million tokens.
- Output tokens include all billable output, including reasoning tokens where applicable. Extra tool fees are per attempt; infrastructure is a monthly fixed estimate.
Assumptions and limits
- OpenRouter catalogue prices are a dated base-rate snapshot; availability, tiers and provider routing can change. Popularity is not a recommendation.
- Audio, images, batch discounts, taxes, cache-write charges, embeddings and provider-specific tier rules are not modeled. Add applicable charges or use custom rates.
- Retries assume the same average tokens and cache mix. Actual bills depend on measured usage and provider billing policies.
FAQs
Does this tool call an AI API?
No. It multiplies your planning assumptions locally and does not make billable inference calls.
Where do model prices come from?
The estimator loads public OpenRouter prices and weekly popularity data through a read-only endpoint. It labels live versus saved snapshot data with its date. Verify current provider tiers and use custom prices for your contract.
What does the cache percentage mean?
It is the expected share of input tokens billed at the cached-input rate, not the percentage of requests with any cache hit. Only eligible cache hits qualify.