# LLM Token Counter — Free Online Tool
https://token-counter.usetool.org/

> LLM Token Counter is a free AI token counter and token calculator for counting tokens across multiple AI models. Supports 26 models from 5 providers: OpenAI (GPT-4o, GPT-4.1, o3, o4-mini), Anthropic (Claude Opus 4.6, Sonnet 4.6, Haiku 4.5), Google (Gemini 2.5 Pro/Flash), Meta (Llama 3.1), and Qwen/Alibaba (Qwen3, Qwen2.5). Shows cost estimation, context window usage, and real-time token counting. No signup, no ads, no server processing — everything runs in your browser.

## Features
- Count tokens for 26 AI models across 5 providers (OpenAI, Anthropic, Google, Meta, Qwen)
- Exact token count for OpenAI models using gpt-tokenizer (o200k_base, cl100k_base encodings)
- Estimated token count for Claude, Gemini, Llama, Qwen (±10-20% accuracy)
- Real-time cost estimation: input and output pricing per model
- Context window usage gauge with color-coded progress bar
- Model selector with grouped dropdown by provider
- Auto-save text and model selection to localStorage
- Dark mode support
- Mobile responsive design (320px+)
- FREE (local) pricing indicator for open-weight models (Qwen, Llama)

## Model Pricing Reference (as of April 2026)
- GPT-4o: $2.50 input / $10.00 output per 1M tokens (128K context)
- GPT-4.1: $2.00 / $8.00 (1M context)
- Claude Opus 4.6: $5.00 / $25.00 (1M context)
- Claude Sonnet 4.6: $3.00 / $15.00 (1M context)
- Claude Haiku 4.5: $1.00 / $5.00 (200K context)
- Gemini 2.5 Pro: $1.25 / $10.00 (1M context)
- Gemini 2.5 Flash: $0.15 / $0.60 (1M context)
- Qwen3 / Llama 3.1: FREE (open-weight, local deployment)

## Use Cases
AI developers estimating API costs, prompt engineers optimizing token usage, product managers budgeting LLM spend, comparing pricing across models, checking if prompts fit within context windows, reducing token count for cost optimization.

## Limitations
Non-OpenAI token counts are estimates (±10-20% accuracy). No batch mode. No file upload. No token visualization. No chat message format counting. All processing is client-side — no API endpoint.

## FAQ
- What is a token? A subword unit, ~4 characters or ~0.75 words in English.
- How many tokens is 1000 words? Approximately 1,333 tokens.
- Is the count exact? Exact for OpenAI models, estimated ±10-20% for others.
- Is my text sent to a server? No, 100% client-side in your browser.

## Technical
Built with Astro 5 (SSG), Preact + Signals (~16KB island), Tailwind CSS v4. Uses gpt-tokenizer for OpenAI encodings loaded via dynamic import. Text never leaves the browser. Part of the UseTool suite at usetool.org.

Last updated: 2026-04-08