AI API Cost Calculator

Pick a model, estimate your usage, and see what it actually costs to run — real per-token pricing for OpenAI, Claude, and Gemini, not a generic industry average.

Not sure? ~1 token ≈ 4 characters or 0.75 words of English text.

Per month

Per 1,000 requests

$0

Per year

$0

This is an estimate using each provider's standard (non-cached, non-batch) rate — actual bills vary with prompt caching, batch processing, tool-use overhead, and pricing changes. Verify current rates directly: OpenAI, Anthropic (Claude), Google (Gemini).

Running an AI agent in production and want help keeping this number predictable? Our AI Agent Monitoring & Evaluation service tracks cost per task alongside accuracy and failure modes.

Building the agent itself, not just estimating what it'll cost to run? See our AI agent development cost and process guide, or try the Tech Stack Recommender.

output by default, which would mangle the JSON's quotes if called per-field instead of once on the whole hash. Takes tool_name: and tool_url: as locals (tool_url computed by the caller via absolute_url(resource.relative_url) — `resource` itself isn't reliably in scope inside a rendered partial, so it's passed in rather than referenced here). Reads data.description directly, same as _head.erb and case_study.erb already do inside partials/layouts. %>