AI token counter and cost calculator
Count tokens and estimate API cost for GPT, Claude, and Gemini. Runs in your browser. Nothing is sent.
- Free
- No signup
- Nothing stored
Paste text, or drop a .txt, .md, or .json file. Nothing is uploaded.
A month is this number times 30.
… tokens on o200k_base · 294 characters · 54 words · — chars/token
Costs use the o200k_base count (0 tokens). Claude, Gemini, DeepSeek, and Mistral are labeled estimate ≈ o200k × 1.0. That is not an official tokenizer for those providers. Prices updated 2026-10-16.
| Claude Sonnet 5.5 estimate ≈ o200k × 1.0 · Pricing source | $0 | $0.005000 | $0.005000 | $15.00 |
| Gemini 4 Argon estimate ≈ o200k × 1.0 · Pricing source Context window is not listed on the pricing page. | $0 | $0.005000 | $0.005000 | $15.00 |
| GPT-6.1 Sol | $0 | $0.005000 | $0.005000 | $15.00 |
How it works
- 01
Paste your prompt or file
Drop a .txt, .md, or .json file, or paste the prompt. The count updates in the browser.
- 02
Pick models and expected output
Choose the models to compare, set the output tokens you expect back, and the cached-input share.
- 03
Compare cost per request and per month
The table prices input and output separately. Copy it as markdown when you want to keep the estimate.
How the count and the price are produced
OpenAI text is counted with o200k_base, and cl100k_base is available for comparison. Claude, Gemini, DeepSeek, and Mistral do not publish a browser tokenizer, so their counts are an estimate equal to the o200k_base count. Use the provider's count-tokens API when the number has to be exact.
Rates were copied on 2026-10-16 from each provider's pricing page, and each row links that page. Input and output are priced separately. A prompt over a model's long-context threshold uses the higher rate for the whole request. Build an MCP server with Next.js is the closest guide to calling these APIs from this stack.
