<aside>

On this page:

</aside>

<aside>

Expanse Docs

🧠 Product Docs

🤖 Prompting Tips

LLM Directory

Prompting Tips & Tricks

Prompt Library

💻 Use Cases

For Work

For Fun

📚 Resources

Prompt Engineering Resources

🍄 About Expanse

Meet the team

</aside>

What is Context Management?

Every message you send to an LLM provides the AI with the information and ‘context’ to generate a relevant response. Each time you send a message, your entire chat history gets sent. Therefore, the longer your conversation, the greater the context. While having more context can provide the AI more information to draw good results, this can also confuse the AI as it has too much information to process before responding to your prompt. ** Context management helps keep the information the AI processes relevant and concise and managing your context well will get you better results while saving you tokens/money.

**What does Context include?

Context Window

The amount of context or information an AI can hold, relies on the size of the Context Window. You can think of an LLM’s Context Window as the equivalent of its working memory. It determines how well it can carry out tasks without forgetting details from earlier in the conversation. It also determines the maximum size of documents or code samples that it can process at once. Therefore, the larger the context window, the more information the AI is able to hold, review and respond to.

<aside> 💡

Context window size is measured in Tokens*.* A common approximation is that one token corresponds to about 0.75 words.

</aside>

<aside> 💡

Get a feel for context size:

Tokens Approximate Text Length LLMs with Comparable Context Windows
256 A short tweet or text message
1,000 A dense email or news article
4,000 ~3 pages (short story, transcript) GPT-3
8,000 ~6 pages (longer article, essay) GPT-3.5
32,000 ~24 pages (novella, detailed report) GPT-4
100,000 ~375 pages (small book, screenplay) Claude 3.5 Sonnet
200,000 ~750 pages (long novel, technical manual) Claude 3.7 Sonnet
1,000,000 ~3,750 pages (large encyclopedia, 500 images) GPT-4o (OpenAI's upcoming model with extended context window)
</aside>

What are Tokens?

Maximum output size

Why should we care?

Managing Context with Expanse

Expanse gives you some great tools to improve your context management:

Skills to Improve your Context Management

1. Use Focus mode

2. Start New Threads or Edit Messages

3. Keep responses concise

4. Use Spaces to stay organized