AI Token Counter
Count tokens, check context limits and estimate API cost before you send a prompt
- Runs in your browser
- Exact counts for o200k_base and cl100k_base
- Context window check
About AI Token Counter
Paste a prompt or drop in a file to count its tokens with the o200k_base or cl100k_base tokenizer, see how much of a model's context window it fills and, where the model's published prices are current, estimate what it would cost to send. If it's too long, trim it to a token budget while keeping a section you mark as protected. Counting runs in your browser. Where a model's tokenizer isn't published, the count is clearly labeled as an estimate.
Why token counts differ between models
Models don’t read characters; they read tokens produced by a tokenizer. The same sentence can be 20 tokens for one model and 26 for another. The counter uses the published o200k_base and cl100k_base tokenizers to give exact counts for models that use them, and labels other counts as estimates.
Planning around the context window
A model’s context window has to hold your prompt, any documents you add and the reply. The counter subtracts the space you reserve for output and a safety buffer, then shows how much room is left. Providers add their own formatting around messages and tools, so leave some margin.
Truncating without losing what matters
If a prompt is too long, set a token budget and choose what to keep: the beginning, the end or both ends. Select a passage in the editor and mark it as protected to keep it regardless. Truncation only removes content; it never rewrites or summarizes.
Further reading
Questions people ask
- Is my prompt sent anywhere?
- No. Counting and truncation run in your browser. Only public model information, such as context sizes and prices, is loaded from MemonaIQ.
- How accurate are the cost estimates?
- They multiply your token counts by the provider’s published prices, which show the date they were last checked. Prices more than 30 days old aren’t used, so a model can show no estimate until its prices are reviewed again. Cached input, tool calls and reasoning tokens can change the real bill.
- Can I add a model that isn’t listed?
- Yes. Add a custom model with its context size, tokenizer and prices. It’s saved only in your browser and labelled as unverified, because MemonaIQ can’t check the figures you enter.