LLM Cost Calculator

It uses published per-token rate cards for OpenAI, Anthropic, and Gemini models to estimate cost based on the volumes you enter. Actual invoices can differ due to prompt length variance, retries, caching behavior, and provider price changes, so treat results as planning estimates.

Input tokens are everything sent to the model: system prompts, conversation history, and user messages. Output tokens are what the model generates in its response. Longer conversation history increases input tokens on every turn.

Several providers offer discounted pricing for input tokens that are repeated across requests, such as a shared system prompt. Enter the average number of input tokens per request that are typically served from cache to reduce the estimate.

Tool calls themselves are usually not billed as a separate line item, but they add extra tokens for the request/response round-trip. This calculator applies an illustrative per-call rate so you can model that overhead; adjust it to zero if your workflow has none.

No. Set the SupportGPT plan to "None" and the agent count to 0 to see the model-only cost. The optional plan and agent fields let you add a flat, illustrative per-agent platform fee on top of the raw token cost.

Yes. After calculating a scenario, click "Add to comparison" to save it to the comparison table below the results. Change the provider, model, or inputs and add again to compare several scenarios side by side.