About this token & cost calculator

Paste a prompt and the tool estimates how many tokens it costs, then prices that request across the current frontier models: GPT-5.6 Sol, Terra and Luna and Claude Fable 5, Opus 5, Sonnet 5 and Haiku 4.5. Set an expected output length and a request count to see what a whole workload costs, and drag the caching slider to model the effect of prompt caching — cached input bills at one tenth of the normal input rate on every model listed here.

The token count is an estimate, not a real BPE tokenization: it uses a character-based approximation (roughly 4 characters per token for Latin script, ~3.1 for Turkish because it is agglutinative, and about 1 token per CJK character), so treat the ±15% band as the working range. Prompts above 272K tokens also trigger the GPT-5.6 long-context surcharge (2× input, 1.5× output), which the tool applies automatically. Everything runs client-side. Related reading: the GPT-5.6 model family explained.