What Is a Context Window? (2026 Guide)

A context window is how much text an AI model can read and remember in a single conversation. In April 2026, GPT-5/5.4 offers 128K tokens, Claude Opus 4.6 offers 1M at standard rates, and Gemini 3 Pro/Ultra stretches to 2M+. Bigger windows unlock long documents and large codebases.

5 min read · Updated Apr 17, 2026

Think of it as short-term memory

A context window is the maximum amount of text a language model can hold in its working memory at one moment. Everything the model reads — your system prompt, every previous message in the conversation, any files you uploaded, and the response it’s generating — has to fit inside this window.

Once you run out of room, older content gets pushed out. The model literally cannot see it anymore.

Why the number is quoted in tokens, not words

Context windows are measured in tokens, not characters or words. A token is a chunk of text (usually 3–4 characters). As a rough rule: 1,000 tokens ≈ 750 English words. So a 128K-token window is about 96,000 words — more than most novels.

How the big three compare in April 2026

ChatGPT (GPT-5, GPT-5.4): 128K tokens on Plus and Pro. Enough for most conversations, code files, and reasonable-length documents.
Claude Opus 4.6: 1M tokens at standard rates on Pro and Max. You can drop in a whole codebase or a full book and ask questions across it.
Gemini 3 Pro: 2M tokens on AI Pro. Gemini 3 Ultra on AI Ultra goes beyond 2M for massive video, audio, and document workloads.

Why this matters for picking a subscription

Context window size directly shapes what you can realistically do on each plan:

Short chats and writing tasks? 128K is fine — ChatGPT Plus or the free tier of any major model handles this easily.
Long-document analysis, legal review, multi-file codebase questions? Bigger is better — look at Claude Pro or Gemini AI Pro.
Video, audio, and multimodal research? Gemini’s 2M window combined with Workspace integration makes it the tool of choice.

If you work with long documents daily, a small window is a hard ceiling no amount of prompt skill can fix.

The catch: “lost in the middle”

Giant context windows sound great, but models often struggle to retrieve details from the exact middle of a very long input. Information at the top and bottom is remembered best. For critical details, repeat them at the end of your prompt, or use retrieval techniques (RAG) to surface only the relevant chunk.

Going deeper

Frequently asked questions

What is a context window in simple terms?

It's the maximum amount of text — prompt plus response — that the model can handle at once. Think of it as the AI's short-term memory for a single conversation.

How big is ChatGPT's context window?

GPT-5 and GPT-5.4 in ChatGPT Plus and Pro use a 128K-token window. That's roughly 96,000 words, or about a 300-page book.

How big is Claude's context window?

Claude Opus 4.6 offers a 1-million-token window at standard rates — roughly 750,000 words, or an entire novel trilogy. That's one of the largest production windows in 2026.

How big is Gemini's context window?

Gemini 3 Pro uses a 2-million-token window, and Gemini 3 Ultra stretches beyond that. Gemini leads on raw context size.

Does a bigger window always mean better answers?

No. Models can lose track of details buried deep in a huge window — this is sometimes called 'lost in the middle.' For most chat, 128K is plenty; giant windows shine only when you actually need to feed in huge documents.

What happens when I exceed the context window?

The provider trims the oldest messages. You'll feel this as the AI 'forgetting' what you said earlier in a long chat.