Cost Simulator
Paste your real prompt and expected response, describe any attached files by type and size — we count the tokens and price the scenario across models. Nothing is uploaded or stored.
Scenario
≈ 0 tokens
No uploads — pick a format and size, and we estimate the extracted text tokens.
No files in this scenario
≈ 0 tokens
Keep the conversation going. Each turn resends the thread so far — priced as cache reads where the model supports prompt caching.
Cached input — optional
A long system prompt, instructions, or knowledge base you resend on every request. With prompt caching it's billed at the cheaper cache-read rate each call instead of full input; the one-time write cost is negligible spread across your monthly volume.
≈ 0 tokens
No cached files
Resent history is billed at the cheap cache-read rate (providers that support prompt caching, with turns inside the cache TTL). Only new tokens pay full input.
How many times this request runs per month
Models to compare
Advanced settings
Loading the tokenizer (a few MB, one time) — showing the fast estimate meanwhile…
Token Count & Cost
Paste a real prompt and expected response — or describe attached files by size — and costs price themselves across the selected models as you type. Text counts follow the selected counting mode; file counts are always size-based estimates.