Skip to main content
tokenmath
Menu

Changelog

What shipped, when.

Pricing refreshes, new models, calculator features, and the occasional design pass. Nothing else.

  1. Editorial redesign

    A visual overhaul plus compare mode, saved scenarios, and live privacy receipts — the redesign moves the calculator from a serviceable utility to something that looks and behaves like a deliberate piece of dev infrastructure.

    • Redesigned every surface in an Editorial Quiet style: layered warm-dark, hero total in parchment gold (mono, tabular), Source Serif 4 accent on editorial pages.
    • Compare mode — toggle on the home calculator to see your prompt scored against all 10 models in a single sortable table.
    • Saved scenarios — keep up to 10 prompt + model + response-length combinations on this browser. Stored only in localStorage; nothing is uploaded.
    • Privacy receipts panel — live counters of outgoing requests, cookies, and localStorage keys so the privacy contract is verifiable by inspection.
    • Cost-split bar and context-window meter inline on the result card.
    • Mobile sticky total bar that surfaces the answer when the result card scrolls out of view.
    • Models index: vendor filters, sort by cost or context window, multi-select for side-by-side comparison.
    • Keyboard shortcuts — press ? for the full list. Includes Linear-style g-then-letter navigation.
    • New wordmark, monogram favicon, and per-model OG cards rebuilt to match.
    • Calibrated empty-state example chips so first-time visitors see a real result in one click.
  2. OpenAI added; brand renamed to tokenmath

    GPT-5 and GPT-4.1 families now use exact tokenization; the rest stay approximate. Project renamed from tokencount to tokenmath.

    • Added GPT-5, GPT-5 Mini, GPT-5 Nano, GPT-4.1, and GPT-4.1 Mini with exact o200k_base tokenization.
    • Brand and domain renamed: tokenmath / tokenmath.dev.
    • Pricing data verified against Anthropic, Google, and OpenAI public pricing pages on 2026-05-09.
    • pSEO routes regenerated with refreshed worked examples per family.
  3. Initial release

    Client-side LLM token counter for Anthropic Claude and Google Gemini. Tokenization in JavaScript on the user’s browser; nothing about the prompt leaves the device.

    • Claude 4.5 Sonnet, Claude 4.5 Haiku, Claude 4.7 Opus, Gemini 2.5 Pro, and Gemini 2.5 Flash supported with cl100k_base / o200k_base approximations.
    • Tiered pricing modeled correctly (Gemini 2.5 Pro doubles input + output above 200k input tokens).
    • Web Worker tokenization for inputs longer than 50,000 characters so the page stays responsive.
    • WebApplication, SoftwareApplication, FAQPage, BreadcrumbList, and Organization JSON-LD on the relevant routes.
    • Programmatic SEO under /token-calculator/[model] with worked examples and per-model FAQ.

Keyboard shortcuts

Press ? any time to reopen this list.

Show this overlay?
Toggle themet
Focus the prompt textarea/
Go to homegh
Go to modelsgm
Go to pricing datagp
Go to changeloggc
Go to aboutga
Close overlays / dialogsEsc