When you work with Copilot in Microsoft 365 apps, the amount of text it can recall from your current conversation directly affects the quality of its responses. This limit is called the context window. If you ask Copilot to summarize a long document or continue a complex chat, it might forget earlier parts of the discussion once the conversation exceeds this window. This article explains the context window size for Copilot, how it works across different Microsoft 365 applications, and what you can do to keep your interactions productive within this limit.
Key Takeaways: Copilot Context Window Limits
- Copilot context window size: Approximately 8,000 tokens, which equals about 6,000 words or 12 pages of text.
- Copilot in Word and Teams: Each app resets the context window for new documents or meetings, so earlier conversations outside the current session are not remembered.
- Best practice for long documents: Break large files into smaller sections and ask Copilot to summarize each part separately to stay within the token limit.
What Is the Context Window for Copilot
Copilot uses a large language model that processes text in units called tokens. A token is roughly four characters or about three-quarters of a word in English. The context window is the maximum number of tokens Copilot can consider at one time when generating a response. For Copilot in Microsoft 365, this window is approximately 8,000 tokens. In practical terms, that equals around 6,000 words or about 12 pages of single-spaced text.
This limit applies to the combined input from your prompt and the document or conversation history. For example, if you paste a 10-page report into Copilot in Word, the model uses nearly all of its context window for the input. It then has very few tokens left for the response. If you then ask a follow-up question, Copilot may drop earlier parts of the conversation to stay within the window.
How the Context Window Affects Copilot in Different Apps
The context window behaves differently depending on the Microsoft 365 application you use:
- Copilot in Word: The context window includes the current document content and the ongoing chat. It does not retain information from previous documents or sessions. If you open a new document, the window resets.
- Copilot in Teams: During a meeting or chat, Copilot remembers the conversation within that specific thread. Once the meeting ends or you start a new chat, the context window clears.
- Copilot in Outlook: For email summaries or drafts, the context window includes the selected email thread. It does not carry over context from other messages.
- Copilot in PowerPoint: The window covers the current presentation slides and any instructions you give. It does not remember previous presentations.
How to Stay Within the Context Window
To get the best results from Copilot, you need to manage the amount of text you input. Follow these steps to avoid exceeding the context window:
- Break long documents into smaller sections
Instead of asking Copilot to summarize a 50-page report, divide it into chapters or sections. Ask for a summary of each section separately. This keeps each prompt well under the 8,000-token limit. - Use concise prompts
Remove unnecessary words from your requests. For example, instead of saying “Can you please take a look at this document and tell me what the main points are,” say “Summarize key points.” Shorter prompts leave more tokens for the document content. - Restart conversations for new topics
If you switch to a different document or subject, start a new Copilot session. This clears the old context window and prevents the model from mixing unrelated information. - Review the conversation history
In Copilot in Word or Teams, you can scroll through the chat history. If you see that the conversation is getting long, consider starting a new thread to reset the window.
Common Misunderstandings About the Context Window
Copilot Forgets Earlier Parts of a Long Chat
If your conversation with Copilot extends beyond the 8,000-token limit, the model drops the oldest parts of the chat to make room for new input. This is not a bug. It is a design limitation of the underlying language model. To avoid this, keep conversations short and focused on one task.
Copilot Does Not Remember Previous Sessions
Each time you open a new document, meeting, or email thread, Copilot starts with a fresh context window. It does not retain any information from earlier sessions. If you need Copilot to reference a previous conversation, you must manually paste the relevant text into the new session.
Large Attachments Can Exceed the Limit
When you attach a file in Outlook or Teams, Copilot reads the file content as part of the input. If the file is longer than 6,000 words, the context window may be filled entirely by the file, leaving no room for your prompt or response. In such cases, Copilot might return an error or a truncated response. Always check the file size before attaching it.
| Item | Copilot in Microsoft 365 | Other Language Models |
|---|---|---|
| Context window size | 8,000 tokens | 4,000 to 32,000 tokens depending on the model |
| Approximate word limit | 6,000 words | 3,000 to 24,000 words |
| Session persistence | Resets per document, meeting, or chat | Varies by platform |
| Best use case | Short documents and focused conversations | Long-form analysis and extended dialogues |
Copilot in Microsoft 365 uses a fixed 8,000-token context window. This is smaller than some other language models that offer 16,000 or 32,000 tokens. For most business tasks like summarizing a few pages or drafting a short email, this limit is sufficient. For large projects, you need to break the work into smaller pieces.
You can now manage Copilot conversations more effectively by keeping prompts short and documents under 6,000 words. If you frequently work with long files, try using Copilot in Word to summarize each chapter individually. For future reference, watch for updates from Microsoft, as the context window size may increase in later versions of Copilot.