Claude explains AI context windows and their limits
Claude’s official documentation defines a context window as the ‘working memory’ of an AI model—all text it can reference during a conversation, including system prompts, past exchanges, and its own reasoning. Measured in tokens, it determines how much of a chat history a model can use at once. For example, Claude Opus 5.5 and Sonnet 5.5 support 1 million tokens, while Sonnet 4.5 handles 200,000…
Key points
- Claude’s context window holds all conversation text, system prompts, and model reasoning up to 1M tokens for Opus 5.5 and Sonnet 5.5
- ‘Context rot’ may reduce accuracy as token limits grow, per Anthropic’s own documentation
- Old chat content can be dropped to prioritize newer information in long conversations
The documentation clarifies common misunderstandings: a larger window doesn’t guarantee better performance due to ‘context rot,’ where accuracy may drop with excessive input. It also distinguishes context windows from training data and notes that older chat content may be dropped to prioritize newer information. Claude’s developer tools suggest resetting conversations or using clear in Claude Code to avoid wasting tokens on irrelevant history.
Model pages: Claude Opus 5.5 → · Claude Sonnet 5.5 →
The story so far
3 episodes →- Claude explains AI context windows and their limitsthis story
[Glossary] What is an AI 'Context Window'? | Why AI forgets or makes mistakes in long conversations
note.com · 4 October 2026
Loading the full article…
This text was published by note.com and written by しらべラボ. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Anthropic prompts users to share voice data to enhance AI models · 5 src
- Reka AI releases research preview of Rho-1 omni-model · 1 src
- OpenAI, Google, and Anthropic launch new AI models in late September · 4 src
- OpenAI launches GPT-6.1 Sol at one-fifth of Astra's price · 43 src
- OpenAI shelves GPT-6.1 Astra after safety concerns about deception and scope · 68 src
Comments
via GitHub Discussions