llm-cost 💰

Count tokens and estimate costs for any LLM model — right in your terminal.

300+ models · Auto-updated prices · Zero dependencies · Beautiful CLI

✨ Features

• 🔢 Accurate token counting — BPE-approximation, ~95% accuracy
• 💵 Real-time pricing — auto-fetches latest prices from OpenRouter on every run
• 🏆 Live Benchmarks — pulls daily ELO ratings directly from LMSYS Chatbot Arena
• 📊 Model Recommendations — smart match by category (code, chat, agents) and cost efficiency
• 🎨 Beautiful interactive UI — colorized TUI with budget calculator and filtering
• 📦 Zero dependencies — pure Node.js, nothing to install
• 🔌 Pipe-friendly — works with cat, echo, stdin
• 📚 Library API — use programmatically in your own projects

Supported Providers

| Provider | Models | |----------|--------| | Anthropic | Claude Opus 4.5, Sonnet 4.5, Haiku 4.5, 3.7 Sonnet, 3.5 Sonnet/Haiku, 3 Opus | | OpenAI | GPT-4.1, GPT-4.1 Mini/Nano, GPT-4o, GPT-4o Mini, o3, o4-mini | | Google | Gemini 2.5 Pro/Flash/Flash-Lite, 2.0 Flash | | Meta | Llama 3.3 70B, 3.1 405B, 3.1 8B | | Mistral | Mistral Large, Small, Codestral | | xAI | Grok 3, Grok 3 Mini | | DeepSeek | DeepSeek R1, Chat V3 | | Cohere | Command R+, Command R | | + 300 more via auto-update from OpenRouter |

🚀 Install

npm install -g llm-cost

Or use without installing:

npx llm-cost

Requires Node.js 18+.

📖 Usage

Interactive Mode (default)

Just run llm-cost with no arguments — perfect for beginners:

llm-cost

It will:

1. Show a welcome screen with model count
2. Let you type/paste your prompt
3. Count tokens in real time
4. Ask for expected output length
5. Show a filterable comparison table

Quick Mode

# Count tokens and show costs for top models
llm-cost "Explain quantum computing in simple terms"

# Specific model
llm-cost "Hello world" -m claude-haiku

# Full comparison across ALL models
llm-cost "Write a 1000-word essay about AI" --compare

# Set expected output tokens
llm-cost "Translate this article" -o 2000

Pipe Mode

# From a file
cat my-prompt.txt | llm-cost

# From echo
echo "Hello, how are you?" | llm-cost

# From clipboard (macOS)
pbpaste | llm-cost

List Models

# All models with prices
llm-cost --list

# Filter by provider
llm-cost --list anthropic
llm-cost --list openai
llm-cost --list google

📦 Programmatic API

const { countTokens, getModels, estimateCost, estimateAll } = require('llm-cost');

// Count tokens
const tokens = countTokens('Hello, world!');
console.log(tokens); // ~4

// Get all models (auto-fetches latest prices)
const models = await getModels();

// Estimate cost for one model
const cost = estimateCost({
  inputTokens: 1000,
  outputTokens: 500,
  model: models[0],
});
console.log(cost); // { inputCost, outputCost, totalCost }

// Compare all models at once
const { results } = await estimateAll({
  text: 'Your prompt here',
  outputTokens: 500,
});
// results sorted cheapest-first

🔄 Auto-Update

Prices are automatically fetched from OpenRouter's public API on every run and cached for 1 hour. If you're offline, built-in fallback prices are used.

No API key required. No account needed.

🛠 CLI Reference

| Option | Description | |--------|-------------| | (no args) | Interactive mode | | "text" | Quick calc for top models | | -m, --model <name> | Calculate for a specific model (partial name OK) | | -o, --output <n> | Expected output tokens (default: auto) | | -c, --compare | Compare ALL models | | -l, --list [provider] | List models with prices | | -h, --help | Show help | | -v, --version | Show version |

🤝 Contributing

Contributions are welcome! Feel free to:

• Add new model providers
• Improve token counting accuracy
• Suggest UI improvements

git clone https://github.com/YOUR_USERNAME/llm-cost.git
cd llm-cost
npm test

📄 License

MIT — use it however you want.