Prompt Token Counter — See Exactly What GPT and Claude Will Charge You
Live token counts, pricing estimates, and context remaining for GPT-6, GPT-5.6, Claude, and Gemini
No signup • Runs in browser • Free
LLM prompts balloon quickly, especially when you paste log files, code snippets, or whole PR descriptions. Copy/pasting into a playground just to see a token count wastes time and leaks sensitive text. The Prompt Token Counter keeps everything client-side with js-tiktoken so your prompt never leaves the browser.
Models covered
Prices are standard API rates in USD per 1 million tokens, checked 23 September 2026. Batch and cached-input discounts are not applied.
| Model | Encoding | Context window | Input / output per 1M tokens |
|---|---|---|---|
| GPT-6 Astra | o200k_base | 272K at standard rate | $10.00 / $50.00 |
| GPT-5.6 Sol | o200k_base | 272K at standard rate | $4.00 / $20.00 (promo through 21 Nov 2026) |
| GPT-5.6 Terra | o200k_base | 272K at standard rate | $2.00 / $12.00 |
| GPT-5.6 Luna | o200k_base | 272K at standard rate | $0.20 / $1.20 |
| Claude Fable 5.1 | o200k_base estimate | 1M tokens | $10.00 / $50.00 |
| Claude Opus 5.5 | o200k_base estimate | 1M tokens | $4.00 / $20.00 |
| Claude Opus 5 | o200k_base estimate | 1M tokens | $5.00 / $25.00 |
| Claude Sonnet 5 | o200k_base estimate | 1M tokens | $2.00 / $10.00 |
| Claude Haiku 4.5 | o200k_base estimate | 200K tokens | $1.00 / $5.00 |
| Gemini 3.8 Flash | o200k_base estimate | 1M tokens | $0.75 / $3.75 (intro, through 31 Dec 2026) |
OpenAI prompts over 272K input tokens are billed at 2× input and 1.5× output for the whole request. Anthropic and Google don't publish their tokenizers, so their counts are estimated with o200k_base. Treat them as a close ballpark, and use the vendor's own token-counting endpoint when you need an exact figure. Check each vendor's pricing page before committing to a budget: OpenAI, Anthropic, Google.
What you see while typing
- Token, character, and word counts update every 100 ms as you edit.
- Estimated spend per model uses current public pricing with a quick link to the vendor's pricing page.
- Context remaining + progress bar shows whether you're safe, approaching the 75% warning zone (amber), or bursting the limit (red).
- Split into chunks proposes how many segments you need at ~80% fill if you exceed a model's window.
- Copy token count instantly copies a model's token total for commit messages, PR templates, or Slack updates.
Everything runs in the browser — no telemetry, no external API calls, and no prompt leakage.
Workflow ideas
- Preflight every long-form prompt. Paste the doc, note the cheapest model that still fits, and flag any wrap-around chunking before handing it to your team.
- Budget multi-model flows. Compare GPT-5.6 Sol and Luna, or Claude Opus 5 and Haiku 4.5, in the same view to decide which step of your pipeline can downshift.
- Keep CI chatbots on budget. Combine the Token Counter with the Diff Checker so that your PR bot can point to exact token costs when reviewers complain.
- Share links with state. The tool URL encodes the textarea content, so you can paste a link in Slack and everyone opens the same tokenizer snapshot instantly.
Need to double-check a mega prompt right now? Open the Prompt Token Counter, paste your text, and keep those GPT invoices predictable.