$0 · 100% browser-side

LLM token counter

Count BPE-calibrated tokens, words, characters, and UTF-8 bytes locally in your browser—and see real per-call and monthly API costs across GPT, Claude, and Gemini models.

Create API Key →
Token boundary preview
You are a financial document parser. Extract every line item from the invoice below and return strict JSON matching the schema { "invoice_id": string, "currency": "USD" | "EUR", "total": number, "items": [{ "sku": string, "qty": number, "unit_price": number }] }. Invoice #INV-2026-4819↵ 1. Cloud GPU cluster (8x H100) - qty 12 hours - $2.15/hr↵ 2. Object storage egress - qty 450 GB - $0.02/GB

Counts & workload

Estimated tokens

121

3.25 chars/token

Characters

393

329 no-spaces

Words

64

Lines / UTF-8 bytes

5 / 393 B

Client-side BPE-calibrated estimate (no text leaves your browser). Exact billing uses the provider tokenizer returned in API usage.

Estimated cost for this prompt across flagship models

Per-call and monthly cost (10,000 calls/mo, 400 output tokens/call) using live KeelRouter rates.

ModelProviderEst. input tokensRate (in / out per 1M)Cost / callEst. monthly
Claude Haiku 5.5Anthropic122$0.12 / $0.61$0.0003$2.59
Claude Opus 5.5Anthropic122$1.2 / $6.1$0.0026$25.86
Claude Sonnet 5.5Anthropic122$0.61 / $3$0.0013$12.74
DeepSeek V4.1 FlashDeepSeek121$0.34 / $1.3$0.0006$5.61
Gemini 3.1 Flash LiteGoogle121$0.3 / $1.8$0.0008$7.56
Gemini 3.5 FlashGoogle121$0.58 / $3.5$0.0015$14.70
Gemini 3.6 FlashGoogle121$0.29 / $1.4$0.0006$5.95
GPT-6.1 SolOpenAI121$0.24 / $1.2$0.0005$5.09

Run this prompt across 100+ models with one API key

Skip juggling separate OpenAI, Anthropic, and Gemini keys. Create one KeelRouter API key and pay only for actual settled tokens.

Frequently asked questions

Does this token counter send my prompt to any server?
No. All tokenization, word/character counting, and cost calculations run 100% locally in your browser with zero network requests.
How accurate are token counts across GPT, Claude, and Gemini?
The counter applies BPE/SentencePiece-calibrated pre-tokenization tuned per model family (o200k_base, cl100k_base, Claude BPE, Gemini SentencePiece, DeepSeek/Qwen). Final API billing always uses the provider’s exact token usage returned in the response.
How is per-call and monthly cost calculated?
Input cost = (input tokens / 1,000,000) × model input rate; output cost = (expected output tokens / 1,000,000) × model output rate, using live KeelRouter catalog pricing.
Can I compare multiple models on the same prompt?
Yes. The comparison table below automatically estimates token counts and per-call + monthly API costs across flagship OpenAI, Anthropic, Gemini, and DeepSeek models.