Hey Product Hunt 👋 I built PromptTokenizer because I got tired of being surprised by LLM API bills. A prompt that looked "short" would somehow consume hundreds of tokens. JSON payloads were larger than expected. Code tokenized differently than prose. And the same prompt could cost dramatically different amounts depending on whether I sent it to GPT, Claude, Llama, Gemini, or another model. Most token counters I found were either: • Limited to a single provider • Hidden behind API keys • Or just gave a number without showing why I wanted a tool that made tokenization visible. So I built PromptTokenizer: 🎨 Visualize every token with color-coded chips 🔢 Count tokens 🤖 Support for GPT models today, with Claude, Gemini, Llama, Mistral, and others planned 💰 Spot expensive patterns before they impact your costs 🔒 100% free and privacy-friendly — everything runs in your browser I use it daily when: • Designing system prompts • Optimizing RAG chunk sizes • Estimating API costs • Comparing models before deployment I'd love your feedback: ❓ Which models should I add next? ❓ What features would make this part of your workflow? ❓ What tokenization edge cases have caught you off guard? Try it here → https://prompt-tokenizer.site/ Thanks for checking it out, and I'm happy to answer any questions in the comments! 🚀

PromptTokenizer

Count tokens for GPT-4, Claude & more — free

Count tokens for GPT-4, Claude & more — free