tokcount

Token counter for all models. GPT, Claude, Gemini. Know your context limits.

Install

npm install -g tokcount

Usage

# Count tokens in text
echo "Hello world" | tokcount

# Count tokens in file
tokcount file.txt
tokcount prompt.md

# Specify model
tokcount file.txt --model gpt-4
tokcount file.txt -m claude
tokcount file.txt -m gemini

# Count directory
tokcount src/

# Show breakdown by file
tokcount src/ --breakdown

# Check against limit
tokcount file.txt --limit 4096

# Compare models
tokcount file.txt --compare

# Output as JSON
tokcount file.txt --json

# Help
tokcount --help

Supported Models

| Model | Context Limit | |-------|---------------| | gpt-4 | 128,000 | | gpt-3.5 | 16,385 | | claude | 200,000 | | claude-3 | 200,000 | | gemini | 1,000,000 | | gemini-pro | 32,000 | | llama | 8,192 | | mistral | 32,000 |

Output

$ tokcount prompt.md --model claude
  tokcount - pnkd.dev
  ───────────────────
  File:    prompt.md
  Model:   claude
  Tokens:  1,234
  Limit:   200,000
  Used:    0.6%

  ✓ Within context limit

$ tokcount file.txt --compare
  MODEL           TOKENS      LIMIT         USED
  --------------------------------------------------
  gpt-4           1,234       128,000       0.9%
  gpt-3.5         1,234       16,385        7.5%
  claude          1,410       200,000       0.7%
  gemini-pro      1,234       32,000        3.8%

License

MIT

pnkd.dev - glitch the system