Basics

Models & limits

Pick the right Claude model and understand what it can and can't hold in mind.

Knowledge cards
Knowledge card
Know What Claude Can't Know
A clear mental model of the boundary between what Claude actually knows — general knowledge up to its training cutoff, plus whatever you give it — and what it cannot know: recent events, live data, and anything private to you or your organisation.
L1 · Basic prompts
Concept
Know What Actually Fits in the Context Window
A working understanding of the context window: the finite budget of tokens Claude can attend to at once, shared across the system prompt, everything you've pasted, the conversation history, and the space left for the reply. It is short-term working memory, not permanent storage.
L2 · Context engineering
Best practice
Match the Claude Model to the Task, Not the Hype
A habit of choosing the Claude model by what the task actually demands — a smaller, faster model for simple high-volume work, the most capable model for hard reasoning — rather than defaulting to the biggest one for everything or the cheapest one to save money.
L1 · Basic prompts
Knowledge card
Start a Fresh Chat When the Thread Drifts
Knowing when to abandon a long, meandering conversation and open a fresh one — carrying over only the context that still matters — instead of pushing on in a thread that has filled with stale or contradictory turns.
L1 · Basic prompts
Claude models and limits

Pick the model for the task

Claude comes in a range of sizes that trade speed against depth. A fast model is right for quick edits, simple questions, and high-volume work where latency matters. A deeper model earns its cost on architecture, hard reasoning, and long-document analysis where getting it right is the point. Defaulting to the strongest model for everything is slow and wasteful; defaulting to the fastest for everything gets you fluent mistakes on the hard tasks. The habit worth building is choosing deliberately based on what the task actually demands.

Know what fits in mind

Claude's context window — everything it can hold in mind at once, including your prompt, the conversation, and any files — is large, but it is not infinite, and relevance matters more than raw size. You can give it long documents, but padding the window with unrelated material dilutes the signal and can degrade the answer. Understanding this limit is what lets you use the big window well: give Claude the material that bears on the task, and leave out the rest.

Frequently asked
Which Claude model should I use?

A fast model for quick edits and high-volume work, a deeper model for architecture, hard reasoning, and long documents. Choose based on what the task demands rather than always using one.

What is Claude's context window?

It is everything Claude can consider at once — your prompt, the conversation, and any files. Claude's window is large, letting it reason over long documents, but relevance still matters more than size.

Does giving Claude more context always help?

No. Padding the context window with unrelated material dilutes the signal and can worsen the answer. Include what bears on the task and leave the rest out.