I built a free, client-side AI API Cost Calculator. Here is the math. published: true

The AI market is moving incredibly fast, and with major players like OpenAI, Google, and Anthropic constantly cutting prices, estimating your API spending is a moving target.

To help developers and startups project their monthly LLM budgets without exposing their proprietary prompts, I built an offline-capable, real-time AI API Cost Calculator.

Try it live: https://tools.kandz.me/ai-cost-calculator

The Core Problem: Words to Tokens 🪙

Large Language Models do not read raw words; they process "tokens" (syllables or character fragments). Standard English text averages a 4/3 ratio—meaning 1,000 tokens are roughly 750 words.

To estimate costs, you must calculate both Input Tokens (your prompt instructions) and Output Tokens (the AI response generation) because output tokens are significantly more expensive to generate.

The Reactive Architecture 💻

The tool uses Angular 18 and reactive Signals to instantly calculate and compare costs across 11 different models (including GPT-4o, Gemini 2.0 Flash, Claude 3.5 Sonnet, and DeepSeek-V3).

By splitting the calculation into pure, non-blocking computed signals, the comparison table renders in real-time as you type, with zero UI lag:

comparisonResults = computed(() => {
  const inTokens = (this.avgInputWords() * 1.33) / 1_000_000;
  const outTokens = (this.avgOutputWords() * 1.33) / 1_000_000;
  const count = this.requestCount();

  return this.models()
    .map((m) => {
      const cost = (inTokens * m.in + outTokens * m.out) * count;
      return {
        ...m,
        totalCost: cost.toFixed(4),
      };
    })
    .sort((a, b) => Number(a.totalCost) - Number(b.totalCost));
});

Total Client-Side Privacy 🛡️

When you are testing system prompts, you are often handling sensitive data like internal APIs, business rules, or private customer records. Sending this data to a remote server for "cost calculation" is a major security flaw.

By running the entire heuristic tokenizer locally on your device, your prompts and variables never leave your browser's temporary memory.

Check out the full comparison table and let me know if you want me to add any specific open-source models: tools.kandz.me/ai-cost-calculator