MONTHLY BUDGET
What will this workflow cost?
Move the sliders to estimate a month of real usage on Claude Sonnet 5.
Estimated monthly spend$7.366
PRIVATE BY DEFAULT · 19 MODELS
Know what your
prompt costs.
Count tokens, check context fit, compare 19 major models, and plan API spend before you ship.
GBuilt for clear decisions — not confusing dashboards.
REGISTRY SNAPSHOT
19 models.
11 providers.
THE CALCULATOR
Start with the text.
Nothing leaves your browser
COUNTING MODE
TRY A SAMPLE
COMPARE MODELS
Same prompt. Different math.
Switch models above or click a card to make it your active calculator.
THE LIVING REGISTRY
More models, less guesswork.
Reference pricing snapshot · verify provider billing before production use.
MORE THAN A COUNTER
Plan the work, not just the prompt.
Small tools that turn a one-off count into a repeat workflow.
PROMPT HEALTH
Is this prompt doing too much?
A quick local check for repetition, context weight, and likely cleanup opportunities.
98/ 100
prompt health
prompt health
- Reasonable prompt size
- Clear section structure
- Local-only analysis
FILE COUNTER
Count a document in seconds.
Drop a .txt, .md, .json, .csv, .py, .js, or .ts file. It stays in your browser.
FROM THE PROMPTCOSTLAB BLOG
Learn, then calculate.
The calculator stays focused. Tutorials, comparisons, and model updates live in their own searchable publication.
Model guidesClaude Fable 5 token limits, pricing, and context fit
A plain-English guide to planning prompts for Anthropic's newest flagship family.
Read article →
ComparisonsKimi K3 vs Claude Sonnet: which prompt costs less?
A practical comparison for long-context chat, RAG, and production assistants.
Read article →
Token basicsWhy the same prompt has different token counts
Tokenizers explained without the textbook: vocabulary, chunks, and hidden request overhead.
Read article →LEARN THE BASICS
Useful answers,
zero jargon.
PromptCostLab is designed for someone who wants a trustworthy answer before they need a full course on tokenization.
Why do models disagree on token count?
Every provider uses its own tokenizer and request formatting. The same sentence can be a different number of tokens on Claude, GPT, Gemini, or Kimi. That is why PromptCostLab labels local counts as estimates.
Does my prompt leave this browser?
Not in Private local mode. Text is counted on your device. Connected exact counting can be added later for structured messages, tools, files, and images when you choose to send data to a provider.
Can I add new models and blog posts?
Yes. The registry and journal are intentionally data-driven: add a model record or article record, and the filters, cards, and comparison surfaces update with it.