Tokens are the basic units that AI uses to process text. Think of them as the "fuel" for AI processing.
Simple rule of thumb:
1 Token ≈ ¾ of a word
100 words ≈ 130 tokens
1,000 words ≈ 1,300 tokens
How Token Consumption Works
📥 Input (what you provide):
Your questions and prompts
Uploaded documents (PDF, Word, Excel, etc.)
Images and screenshots
Audio files (after transcription)
Previous conversation history
📤 Output (AI responses):
All AI answers
Summaries and analyses
Generated texts and lists
⚠️ Important (Cloud version): Input and output are weighted equally - both count fully toward your token consumption.
Examples of Token Consumption
Typical amounts:
Short question: "What's in the document?" ≈ 8 tokens
Normal answer: 200 words ≈ 270 tokens
A4 page of text: ≈ 500-800 tokens
Excel spreadsheet: 100 rows × 5 columns ≈ 500+ tokens
Particularly token-intensive content:
📊 Excel files:
Every single cell is counted
Even empty cells can consume tokens
Tip: Remove empty rows/columns before upload
🖼️ Scanned PDFs (OCR):
Text must first be recognized (costs extra tokens)
Then normal text processing
Tip: Use PDFs with selectable text
🎵 Audio files:
First transcription to text (costs tokens)
Then text processing (costs tokens again)
Example: 10 min audio ≈ 1,500 words ≈ 2,000 tokens
Optimizing Token Consumption
Efficient usage:
✅ Ask precise questions:
Instead of: "Tell me everything about the document"
Better: "What are the 3 main points in the document?"
✅ Upload relevant content:
Only needed pages/sections
Remove empty Excel rows
Focus on important information
✅ Structured requests:
One clear question per chat
Specific rather than general requests
Avoid token traps:
❌ Very large Excel files with many empty cells
❌ Complete documents when only parts are needed
❌ Repeated uploads of the same content
❌ Very long, unstructured questions
Cloud vs. On-Premises
Cloud version (standard):
Token counting active - input + output is calculated
Fair billing based on actual consumption
Transparent costs based on usage
On-Premises hosting:
No token calculation - you pay for hardware/license
Unlimited usage within your infrastructure
Own control over data processing
Frequently Asked Questions
How many tokens have I used?
Your current consumption is displayed in the dashboard.
Are tokens still counted for errors?
Yes, even failed processing consumes tokens.
Can I predict my token consumption?
Roughly yes - use the guidelines above as orientation.
What happens when my tokens are used up?
Depending on your plan, additional tokens will be charged or usage will be paused.
Are deleted chats refunded?
No, already consumed tokens remain consumed.
Practical Tips
Before uploading:
✅ Reduce file size - only relevant content
✅ Optimize Excel - remove empty areas
✅ PDF quality - selectable text instead of scan
During usage:
✅ Ask specific questions
✅ Proceed step by step instead of all at once
✅ Reuse results instead of regenerating
🧠 In short:
👉 Tokens = AI "fuel"
👉 Input = Your data (questions + documents)
👉 Output = AI responses
👉 Cloud: Input = Output (same counting)
Tip: Think of tokens like your data volume - conscious use saves costs and improves efficiency.
Questions about token consumption or optimization? → Contact support
