About LLM Counter

Free, fast, single-purpose tools for people working with AI models.

LLM Counter answers three practical questions: how many tokens is this text, can my hardware run this model, and what will an API workload cost. Each tool is deliberately simple — one question, one clear answer — and everything runs in your browser, so nothing you type is ever uploaded. We don't sell anything, there are no accounts, and no "book a demo" agenda. The site is supported by advertising.

How the numbers are verified

Every model spec and price on this site lives in a single shared registry — currently covering 47 models — and every figure in it traces to a primary source: the provider's official pricing page, or the model's published configuration on Hugging Face. We never take numbers from blogs, news coverage or other calculators, and each registry entry records the source URL and the date it was last checked.

That policy is enforced, not just intended: an automated test suite validates the registry before every deploy, and a staleness check fails the build outright if any entry goes more than sixty days without re-verification. Where a number is an estimate rather than exact — Claude and Gemini token counts, for example — it is labelled as one everywhere it appears. The methodology page documents every formula, its accuracy, and where its inputs come from.

How often things update

The AI model landscape moves monthly, so the registry is re-verified on a monthly cycle — the last full pass was 2026-09-03 — and new models are added within days of release. Every page footer shows the date its data was last verified, so you never have to guess whether a price is current.

Who publishes it

LLM Counter is published by Cactus & Cubby Marketplace (ABN 28 316 231 089), an Australian business that builds and maintains free single-purpose web tools. If you spot an outdated price or a model we're missing, it will usually already be on the next monthly pass — but the fastest fix is the freshest source: every figure's origin is listed on the methodology page.