llm-bill

Changelog

What changed

Price changes, re-measurements, method updates and new comparisons, newest first. Each entry says what it changed and links the pages it affected. Dates are when the change was published.

  1. · Launch

    First comparison published

    First published comparison of Claude Haiku 5.5 and GPT-6 Luna: tokenizer coefficients by content type, structured-output overhead, three measured tasks (ticket classification, invoice extraction, Japanese business email), blind judging of the email task, and the cost calculator. Measurement dates, sample sizes and sources are shown on each page and on the method page.

    Pages: Calculator · Report · 日本語レポート · Method & data