Why the count is an estimate
Provider tokenizers split text differently, and newer model families may change tokenization. Use the provider's API usage field or official tokenizer for exact billing reconciliation.
Files are read locally by your browser. This version intentionally supports text-based formats only and does not pretend to parse PDF or DOCX.
Nothing leaves your device.
Provider tokenizers split text differently, and newer model families may change tokenization. Use the provider's API usage field or official tokenizer for exact billing reconciliation.
Reliable document cost depends on extraction, OCR, images and layout. This tool does not claim to support those formats until a proper local parser is implemented.