Claude Context Window

The context window of every current Claude model is 1M tokens, roughly 555,000 English words or 2.5 million characters, and each can write up to 128K tokens in a single response (Models overview, Anthropic docs, read Nov 6, 2026). On the API the full 1M is the default, with no beta header and no long-context surcharge, except that Claude Haiku 5.5 charges a higher rate for prompts over 100,000 tokens. In the Claude apps, paid plans get up to 1M depending on the model, and Anthropic's Help Center says other models get 500K or 200K (How large is the context window on paid Claude plans?, Claude Help Center, read Nov 6, 2026).

Quick facts

Sources: Models overview, Context windows and Pricing, Anthropic docs, read Nov 6, 2026; Plans and pricing, Claude, read Nov 6, 2026.

What is a context window?

0:00 The context window is everything Claude can read at once in one request: the system prompt, the whole conversation so far, attached files and images, tool definitions and results, and the answer it is writing. It is measured in tokens, the pieces of text a model reads, each usually a word or part of one. Once a conversation would pass the window, something has to give: older turns are dropped, summarized or cleared.

How many words fit depends on the tokenizer. Models from Claude Opus 4.7 on use a newer one that turns the same text into about 30% more tokens, so 1M tokens holds about 555,000 words, against about 750,000 on earlier models; 200K tokens is about 150,000 words on the older tokenizer (Models overview; Pricing, read Nov 6, 2026). Anthropic's launch posts gave the increase as 1.0 to 1.35 times, depending on the text.

Context window by model (API)

Sources: Context windows and Thinking, Anthropic docs, read Nov 6, 2026. The 300K Batch ceiling needs the beta header output-300k-2026-03-24 and also covers Opus 5, Sonnet 5, Opus 4.6 to 4.8 and Sonnet 4.6, but not Fable or Mythos models. A request whose input plus max_tokens exceeds the window is accepted on Claude 4.5 and newer models and stops with stop_reason: "model_context_window_exceeded".

Context window in the Claude apps

The Help Center's table for paid plans (Pro, Max, Ultra, Team and Enterprise), read Nov 6, 2026:

Source: How large is the context window on paid Claude plans?, Claude Help Center, read Nov 6, 2026. In Cowork, Sonnet 5 compacts at 500K. In Claude Code, Opus 4.6's 1M needs usage credits on Pro, and Sonnet 4.6's needs them on every plan but usage-based Enterprise.

That article covers only paid plans. The pricing page lists "Up to 1M, varies by model" for every individual plan, Free included (Plans and pricing, read Nov 6, 2026), while the older Team plan article still says 200K; the pricing page is the newer of the two. Anthropic does not give a separate figure for the Free plan's Sonnet 5.5 and Haiku 5.5.

What happens when a chat gets too long?

On paid plans with code execution on, Claude summarizes earlier messages as a conversation nears the limit, so most chats can go on indefinitely; the reader may see Claude "organizing its thoughts". The full history is kept, and the summarizing uses more of the plan's usage limits (How large is the context window on paid Claude plans?). Anthropic added this compaction on Nov 24, 2025. Very large first messages can still hit the limit.

2:30 To make room, the Help Center suggests using Projects, which load only the relevant parts of their files, keeping project instructions short, and turning off tools and Connectors that are not needed.

1:48 Bigger is not automatically better. Anthropic's docs warn of "context rot": as the token count grows, accuracy and recall fall (Context windows, read Nov 6, 2026). Starting a fresh chat for a new topic often gives better answers than carrying a long one on.

Compaction in Claude Code and the API

0:36 Claude Code compacts automatically at about 967K tokens on 1M models; /compact summarizes on demand, /autocompact sets the threshold, and CLAUDE_CODE_DISABLE_1M_CONTEXT=1 caps the window at 200K (Model configuration, Claude Code docs, read Nov 6, 2026).

On the API, server-side compaction (beta) replaces older turns with a summary Claude writes. It comes in two forms: on demand (compact-2026-09-04, which the docs recommend wherever available) and automatic at a token threshold (compact-2026-01-12, from 150,000 input tokens by default) (Compaction, Anthropic docs, read Nov 6, 2026). Context editing can also clear old tool results and thinking blocks.

How Claude's context window grew

• 9K to 100K tokens on May 11, 2023.
• 200K with Claude 2.1 on Nov 21, 2023, kept for the whole Claude 3 family.
• 1M in beta with Claude Opus 4.6 (Feb 5, 2026) and Sonnet 4.6 (Feb 17, 2026), out of beta at the standard price on Mar 13, 2026.
• 1M as the default on every current model since the Claude 5 family.

Frequently asked questions

How big is Claude's context window? Every current Claude model, including Claude Opus 5.5 and Sonnet 5.5, has a 1M-token context window on the API, about 555,000 words. In the Claude apps it is up to 1M on paid plans depending on the model.

What is Claude's output token limit? Current Claude models can write up to 128K tokens in one API response, or 300K through the Batch API with a beta header on models such as Opus 5.5, Sonnet 5.5 and Haiku 5.5.

Does Claude charge more for long prompts? Not on most models: Claude 4.6 and later models bill a 900K-token request at the same rate as a 9K one. Claude Haiku 5.5 is the exception, costing $0.50 instead of $0.10 per million input tokens above 100,000 tokens.

What happens when a Claude chat hits the context limit? On paid plans with code execution on, Claude summarizes earlier messages automatically and keeps going, at some cost to usage limits. Without it, or with a very large first message, Claude says the conversation is too long and a new chat is needed.

How many pages fit in Claude's context window? Anthropic says 200K tokens is about 500 pages of text; at roughly 555,000 words, 1M tokens on current Claude models holds a few thousand pages. A single API request can also include up to 600 images or PDF pages.

Sources

• Models overview, Anthropic docs, read Nov 6, 2026
• Context windows, Anthropic docs, read Nov 6, 2026
• Pricing, Anthropic docs, read Nov 6, 2026
• Thinking, Anthropic docs, read Nov 6, 2026
• Compaction, Anthropic docs, read Nov 6, 2026
• Release notes, Anthropic docs, read Nov 6, 2026
• How large is the context window on paid Claude plans?, Claude Help Center, read Nov 6, 2026
• Plans and pricing, Claude, read Nov 6, 2026
• Model configuration, Claude Code docs, read Nov 6, 2026
• Introducing 100K Context Windows, Anthropic, May 11, 2023
• Introducing Claude 2.1, Anthropic, Nov 21, 2023
• Context Management in Claude Code, Claude, May 18, 2026 (video)