菜单

发车步骤 · 扩量前先算成本

ChatGPT Token 计数与费用估算

粗估 ChatGPT / Claude / Gemini 的 Token 与花费。可编辑单价，无需 API Key — 内容不离开浏览器。工具按钮暂为英文。

Method: LLM token cost estimation · Count tokens without an API key

Empty? Load a ticket-summarizer launch scenario (Luna · 100k calls/mo · 10% retries), or paste your own prompt below.

  1. 1Paste or load example
  2. 2Pick a model
  3. 3Read $/call
Model Input / 1M Output / 1M Use

Approximate client-side token estimate (chars÷N, not tiktoken). Default rates are public list shapes verified 2026-09-22 against OpenAI, Anthropic, Google Gemini — re-check before you budget. Rates are editable; Save stores them in this browser only. Gemini 3.8 Flash defaults use introductory paid standard ($0.75 / $3.75 per 1M) through 2026-12-31; list price doubles 2027-01-01 per Google’s pricing page — do not bake the promo into steady-state forecasts.

Empty — load the launch example or paste a prompt to see $/call.

Input tokens—
Output tokens—
Cost / call—
Monthly (w/ retries)—

FAQ

Is this tiktoken / the OpenAI tokenizer?

No. This page uses a client-side chars÷N estimate (default ~4 chars per token), not OpenAI tiktoken or any vendor tokenizer. Use it for planning ranges, then verify with the vendor tokenizer when you need exact billing. See Local vs vendor tokenizer and Count tokens without an API key.

Can I use this as a ChatGPT / OpenAI token counter without an API key?

Yes for rough counts. Paste text, keep or tune chars-per-token, and read the input-token estimate. Everything runs in your browser — prompts are not uploaded. It is still an estimate, not a Tiktoken/API count.

How do I use this as an LLM / OpenAI cost calculator?

Paste sample prompt text, set expected output tokens, calls/month, and retry rate. Pick or edit USD-per-1M input/output rates in the table (defaults are public list shapes as of 2026-09-22 — verify vendor pricing pages before you budget). Estimate shows cost/call and monthly spend with retries. Save rates stores edits in this browser only.

Why are Claude and Gemini on the same page as ChatGPT?

One client-side estimator covers GPT, Claude, and Gemini so you can compare rough monthly spend without a doorway farm of near-duplicate token-counter URLs. Switch models in the rates table or use Show all models.

Looking for tiktoken online without installing Python?

This tool is a browser-side chars÷N estimate — not OpenAI tiktoken — so you can rough-count tokens and $/call without installing Python or calling an API. For exact vendor billing, use the official tokenizer; for an offline playbook see Count tokens without an API key. Prefer Download CSV or Download .md when clipboard is blocked.

Related playbooks

Guides are curated playbooks tied to this tool — not volume SEO filler.