When working with CoAuthor, it's essential to understand the distinction between file size limits and context limits—two factors that affect how your content is handled by the system and interpreted by the AI.
File Size vs. Context Window
File Size Limit: CoAuthor supports uploads up to 2GB. This determines how much data can be stored or transferred into the system.
Context Window: The GPT model powering CoAuthor has a context window of 128,000 tokens, which translates to approximately 96,000 words (based on the rule of thumb: 0.75 tokens per word). This defines how much content the model can actively process and understand at one time.
Why Context Matters More Than Size
Even if a file is under the 2GB upload limit, it may still exceed what the model can interpret if the text content surpasses the token limit. This is especially relevant for large documents with extensive text, such as technical manuals or research papers.
Practical Considerations
Word files over 20MB may sometimes be too large to process efficiently, depending on their content.
For example: Word files over 20MB may sometimes be too large to process efficiently, depending on their content. For example:
Uploading 60 MB PDF of TFL tables as a source, may result in a "Timeout" after several minutes.
This issue was not solely due to the file size limit (60 MB), but also because the large PDF was expected to be used as a source by CoAuthor.
Smaller files (7-10 MB) can take several minutes to convert, emphasizing the importance of considering both file size and content when preparing documents for CoAuthor.
Summary
Upload size determines what can be stored.
Token count determines what can be interpreted.
Always consider both when preparing files for CoAuthor to ensure smooth processing and optimal AI performance.