# Token Scope

> A free token counter and LLM API cost calculator at https://token-scope.skillsafe.ai/ . It runs entirely in the browser: no account, no model call, and text or files never leave the page.

## What it does

- Counts tokens exactly with OpenAI's o200k_base (GPT-4o, GPT-4.1, GPT-5, o1/o3/o4, gpt-oss) and cl100k_base (GPT-4, GPT-3.5, text-embedding-3) vocabularies. The tokenizer is this app's own byte-level BPE, checked token-for-token against OpenAI's tiktoken.
- Gives labelled estimates with a low-high range for Claude, Llama 3.x/4, Gemini/Gemma, Mistral and other open models (Qwen, DeepSeek, GLM, Kimi). These are a ratio applied to an exact count, not the real tokenizer. For exact Claude or Gemini counts, use the provider's free token-counting endpoint.
- Shows characters, words, lines, UTF-8 bytes, characters per token and words per token.
- Colours every token boundary, optionally shows whitespace and token ids, and copies the id list.
- Accepts pasted text and dropped files: TXT, Markdown, code, JSON, CSV, HTML, RTF, PDF (text layer) and DOCX, several at once, with per-file counts.
- Checks fit against 8K, 32K, 128K, 200K, 400K, 1M and 2M context windows (plus a custom one) with tokens reserved for the reply.
- Prices input and reply for 49 text models (Anthropic, OpenAI, and open models served by Cloudflare Workers AI) from a dated price list, with a number of calls, a cached-input share and a CSV export.

## Prices

Rates are provider list prices per million tokens recorded in the SkillSafe model catalogue on 2026-10-31. A second view shows the SkillSafe billed rate (list + 10%), which is what a run costs on SkillSafe with credits and no API key. Prices change; check the provider before budgeting.

## Not included

Chat APIs add a few tokens per message for roles and formatting; image, audio and tool-schema tokens are not counted. Special markers such as <|endoftext|> are counted as ordinary text.

## Related

- Token Desk (https://token-desk.skillsafe.ai/): how conversation cost grows turn by turn.
- Context Lens (https://context-lens.skillsafe.ai/): how much context an agent's setup files use.