Prompt Token Cost Calculator

Paste a prompt and see exactly what it costs to send through every major LLM — uncached and cached. Side-by-side comparison ranks the cheapest option for your specific text. Useful for picking a model when cost matters more than capability.

How to use the Prompt Token Cost Calculator

Paste your prompt, set the expected output token count, and pick a monthly call volume. The tool counts tokens with each provider's tokenizer (or a calibrated heuristic where the tokenizer isn't available client-side), then multiplies by current per-million-token pricing. The result is a sorted table — cheapest first.

How to interpret the comparison

Per-token price isn't the only factor. The tokenizer matters too — Gemini and GPT-5/4o (o200k_base) typically produce 5-15% fewer tokens for the same English text than Claude or GPT-3.5 (cl100k_base). For non-Latin scripts the difference can be 2-3x. So the cheapest-per-token-listed model isn't always the cheapest-per-call for your prompt.

Cached pricing applies only to input tokens that match a stable prefix you've sent before. If your prompt is mostly a stable system message with a small per-request suffix, the cached rate dominates and your effective cost drops 90% on supported providers. The comparison shows both columns so you can see the gap.

Output token cost is included in the totals. Output tokens are 2-5x more expensive than input on every provider, so prompt design that constrains output length ("answer in one sentence") is often the biggest single cost lever.