Claude Context Window Size (2026): 1M Tokens on Opus 4.8 and Sonnet 5

Every Claude Model's Context Window

Current Claude models split into two context tiers. Opus 4.8, Sonnet 5, and Fable 5 carry a 1M-token window; Haiku 4.5 carries 200K. The 1M window is generally available at standard per-token pricing, so the differences that remain are max output, price, and the 200K vs 1M split.

Sonnet 5 introductory pricing of $2/MTok input and $10/MTok output runs through August 31, 2026, after which it lists at $3/$15. The 1M window no longer requires a beta header or a usage tier. On the Message Batches API, Opus 4.8, Opus 4.7/4.6, Sonnet 5, and Sonnet 4.6 support up to 300K output tokens via the output-300k-2026-03-24 beta header. Figures are from Anthropic's model docs (platform.claude.com), July 2026.

Context Window vs Effective Context

A context window is how many tokens you can send. Effective context is how many of those tokens the model uses reliably. They are not the same number.

Chroma published one of the most comprehensive studies of this gap in July 2025, testing 18 models across controlled experiments. The headline finding: performance degrades at every context length increment, not just near the limit. A model with a 1M token window still exhibits measurable degradation at 50K tokens.

Adobe Research's NoLiMa benchmark (arXiv 2502.05167, ICML 2025) sharpens the point. It extends needle-in-a-haystack with needles that share almost no words with the question, forcing the model to infer the link rather than string-match. Under that harder test, 11 of 13 models that advertise at least 128K context dropped below 50 percent of their short-context baseline by 32K tokens, and GPT-4o fell from 99.3 percent to 69.7 percent. NoLiMa defines effective length as the longest context where a model keeps at least 85 percent of its base score, which lands far below the advertised window for every model tested.

The Lost-in-the-Middle Problem

Research from Stanford and others established the U-shaped attention curve. LLMs attend most strongly to tokens at the beginning and end of the context. Information in the middle gets deprioritized. This pattern holds across all transformer-based architectures.

Chroma's study added a counterintuitive finding: models performed better on shuffled, incoherent contexts than logically structured ones across all 18 models tested. The attention mechanism behaves differently with structural patterns, suggesting that coherent documents may actually make retrieval harder, not easier.

NVIDIA's RULER benchmark puts effective context at 50-65% of advertised capacity for most models. A model advertising 200K tokens typically becomes unreliable around 130K. Claude Opus 4.6 is a notable exception: its 76% score on MRCR v2 (finding 8 needles across 1M tokens) represents what Anthropic calls "a qualitative shift" in usable context, up from 18.5% for its predecessor.

How Claude Compares to GPT-4.1, Gemini, and Others

Claude no longer trails on raw window size. Opus 4.8 and Sonnet 5 ship 1M tokens, matching GPT-4.1's 1M and sitting under Gemini 2.5 Pro (1M, 2M coming) and Llama 4 Scout's claimed 10M. Haiku 4.5 stays at 200K as the fast tier.

On effective utilization, the picture separates from the spec sheet. Chroma's research found that Claude models "decay the slowest overall," while GPT models were "more erratic with random mistakes and outright refusals," and Gemini "starts to mess up earlier with wild variations."

GPT-4.1 claims 100% needle accuracy at 900K tokens, but needle-in-a-haystack tests measure retrieval of a single planted fact, not reasoning over distributed information. Chroma's study found GPT models had the highest hallucination rates when distractors were present and a 2.55% task refusal rate. Gemini 2.5 Pro reports 99.7% recall at 1M tokens on Google's own benchmark, but Gemini 3 Pro dropped to 26.3% at 1M on its own model evaluation card.

The numbers reveal a pattern: models perform well on the benchmarks their makers design. Independent evaluations like Chroma's paint a less optimistic picture for every model.

Prompt Caching: 90% Cost Reduction

Prompt caching is Anthropic's mechanism for reusing processed context across API calls. Instead of reprocessing the same system prompt, documents, or conversation history on every request, the API reads from cache at a fraction of the cost.

The economics are straightforward. A cache write costs 1.25x base input price for a 5-minute TTL, or 2x for a 1-hour TTL. Cache reads cost 0.1x base input price. For Sonnet 5 at $3/MTok list input, that means a cache read costs $0.30/MTok, 10x cheaper than reprocessing. The break-even is two API calls for 5-minute caching.

For a concrete example: a 100K token system prompt with documents drops from 11.5 seconds to 2.4 seconds with caching. An application making 100 API calls per hour with a shared 50K token system prompt saves roughly $14.25/hour on Sonnet 5 list pricing versus reprocessing.

Extended Thinking and Context

Extended thinking lets Claude reason through problems before responding. The thinking tokens are billed as output but count toward the context window during the current turn. The critical detail: previous thinking blocks are automatically stripped from context in subsequent turns.

This means a multi-turn conversation where Claude thinks for 20K tokens per turn does not accumulate 100K tokens of thinking across 5 turns. Each turn's thinking is discarded before the next, so reasoning does not eat your context window over time.

Opus 4.8 and Sonnet 5 support 128K max output tokens, double the older 64K limit, and up to 300K on the batch API via the output-300k-2026-03-24 beta header. This gives extended thinking more room to reason through complex problems without hitting output limits. The thinking budget is set via the budget_tokens parameter, and with adaptive thinking, Claude dynamically decides how much thinking each request needs.

How Much Code Fits in 200K Tokens

Token-to-code ratios vary by language. Python and JavaScript average about 2.5-3.5 tokens per line of code. Verbose languages like Java run higher. Comments and docstrings add overhead. Here are practical estimates:

A typical React project with 200 files averaging 150 lines (30,000 total lines) fits comfortably in 200K tokens. A 100,000+ line enterprise codebase does not. But fitting everything into context is usually the wrong approach.

Cognition (makers of Devin) measured that their AI coding agent spent 60% of its time searching for code before making changes. The bottleneck was not context window size. It was finding the right code to put in context.

Context Engineering > Context Size

The context window arms race, from 4K to 200K to 1M to 10M, misses the point. Chroma's research demonstrated that performance degrades at every length increment across all 18 models tested. Bigger windows give you capacity, not quality.

Context engineering is the practice of curating what goes into the window: selecting relevant information, removing noise, structuring content for attention patterns, and managing token budgets across multi-turn interactions. Anthropic's own documentation states that Claude "achieves state-of-the-art results on long-context retrieval benchmarks, but these gains depend on what's in context, not just how much fits."

Practical strategies

The models with the biggest context windows are not always the best choice. A 200K window filled with precisely relevant code outperforms a 1M window filled with an entire repository. The tooling that selects and shrinks what goes into context determines outcomes more than the window itself. Morph Compact handles the shrinking; if you want a raw 1M-token window on an open-weight model, Morph also serves GLM-5.2 and DeepSeek V4 Flash at 1M context through one OpenAI-compatible API.

Frequently Asked Questions

What is Claude's context window size?

Claude Opus 4.8, Sonnet 5, and Fable 5 have a 1 million token context window; Haiku 4.5 has 200,000 tokens. The 1M window is generally available at standard per-token pricing as of March 2026, not a gated beta. 1M tokens holds roughly 555,000 words; 200K holds roughly 150,000 words or 500 pages. Source: Anthropic model docs (platform.claude.com), July 2026.

How does Claude's context window compare to GPT-4.1 and Gemini?

Opus 4.8 and Sonnet 5 have a 1M window, matching GPT-4.1. Gemini 2.5 Pro has 1M with 2M coming. Raw size no longer separates them, so effective context is the deciding metric: NVIDIA's RULER puts usable context at 50-65 percent of advertised, and Adobe's NoLiMa found most models drop below half their short-context score by 32K tokens. Chroma's research found Claude models decay the slowest.

What is prompt caching and how much does it save?

Prompt caching lets you mark context for reuse across API calls. Cache reads cost 10% of standard input price, saving up to 90% on costs and 85% on latency. A 100K token prompt drops from 11.5s to 2.4s with caching. The break-even is two API calls for 5-minute caching, one call for any frequently used context.

Does Claude's performance degrade with longer contexts?

Yes. All LLMs exhibit context rot. The lost-in-the-middle effect means information in the center of context gets deprioritized. Claude Opus 4.6 made significant progress, scoring 76% on MRCR v2 vs. 18.5% for its predecessor, but degradation still occurs at every length increment. The practical response is context engineering: sending relevant content rather than maximizing volume.

How much code fits in Claude's 200K token context window?

Approximately 50,000-70,000 lines of code. A medium-sized project (200 files, 150 lines each) fits comfortably. A large enterprise codebase (100K+ lines) does not. For large codebases, semantic search tools that find relevant code before context loading produce better results than trying to fit everything in.

What is the pricing for Claude's 1M token context window?

As of March 2026, Anthropic removed the long-context surcharge. The 1M window bills at standard per-token rates across its full length: Opus 4.8 at $5/$25 per MTok, Sonnet 5 at $3/$15 ($2/$10 introductory through August 31, 2026). A 900K-token request costs the same per-token rate as a 9K one. The old 2x-input, 1.5x-output premium above 200K no longer applies. Prompt caching multipliers still apply on top.

Related Reading

• Context Rot: Why LLMs Get Worse with More Input
• LLM Context Window Comparison (2026)
• LLM Context Windows Explained
• Context Engineering for AI Agents
• Morph Compact: Context Compression at 33,000 tok/s
• GLM-5.2: 1M-Token Open-Weight Model on Morph
• DeepSeek V4 Flash: 1M Context on Morph
• WarpGrep: Semantic Codebase Search