Pick a model, enter what a typical run consumes and how often it runs. Get cost per run, per successful run, per customer and per month, from published list prices, instantly.
List prices per 1M tokens, verified August 2026. Retries, loops and tool calls are not in these numbers, which is exactly why estimates and invoices disagree.
Real runs retry, loop, and grow their prompts. AgentPing prices every actual run from its model and token counts, tagged by agent, customer and feature, so the number on this page becomes a number you can bill against.
Multiply the tokens each run consumes by the model's per-token price, separately for input and output tokens since they are priced differently, then multiply by run volume. Cost per run = (input tokens x input price + output tokens x output price). Monthly cost = cost per run x runs per day x 30. This calculator does that from published list prices for the major models.
Total spend divided by the number of runs that actually achieved their goal, not just completed. If your agent costs $0.05 a run but only 80% of runs succeed, the real unit cost is $0.0625, because the failed runs still burned tokens. It is usually the most honest number for agent unit economics, and the one that changes pricing decisions.
Retries, loops, tool calls and prompt growth all add tokens the estimate cannot see, and cached input pricing can reduce them. Estimates based on averages routinely miss by 2-5x in either direction. The only accurate number comes from measuring each real run, which is what AgentPing does: every run reports its model and token counts, and the cost is priced server-side per run, per customer and per feature.
The current flagship and workhorse models from OpenAI (gpt-4o, gpt-4o-mini, gpt-4.1), Anthropic (Claude Opus, Sonnet and Haiku), and Google (Gemini 2.5 Pro and Flash), at published list prices per million tokens. Prices were last verified August 2026; providers change them, so check yours before making commitments.
Yes, free with no signup. It runs entirely in your browser. If you want the real numbers instead of estimates, AgentPing has a free plan with no card required.
Why cost per successful run is the number that matters → Attributing LLM cost per customer →