Claude 1M Token Context Generally Available — Process 600-Page PDFs in One Go
Anthropic has launched 1 million token context windows for Claude Opus 4.6 and Sonnet 4.6. Process 600-page PDFs and 600 images at once with no extra charge, achieving #1 long-context retrieval accuracy (78.3%).
Claude 1M Token Context — An AI That Reads and Remembers 3 Entire Novels at Once
Ever experienced AI forgetting the beginning of a long document? Anthropic has tackled this problem head-on. Claude Opus 4.6 and Sonnet 4.6 can now read and process 1 million token context windows (approximately 750,000 words — 3 novels, 1,500 A4 pages) in a single session.
Generally available since March 13, this update has moved past beta and is ready for everyone to use immediately. With 1,179 points and 502 comments on Hacker News, developer community interest is intense.
Claude 1M Token Pricing — No Extra Charge for Long Documents
The most notable aspect is the pricing. Most AI services charge extra for long documents, but Claude announced a "No long-context premium" policy.
Simply put, whether you analyze 2 pages or 1,500 pages, the per-page cost is identical.
Claude API Pricing (for developers)
- Opus 4.6: Input $5 / Output $25 (per million tokens)
- Sonnet 4.6: Input $3 / Output $15 (per million tokens)
Individual users on claude.ai can access via Pro/Team/Enterprise subscriptions.
Process Up to 600 Pages of PDFs and 600 Images at Once
It's not just about reading more text. This update increased the limit for images and PDF pages from 100 to 600 — a 6x increase.
In practical terms:
- Feed in hundreds of pages of contracts and ask "find unfavorable terms in Section 3, Clause 2"
- Upload 5 competitor reports simultaneously for comparative analysis
- Submit hundreds of design mockups for review and feedback at once
If document analysis with AI is new to you, start with the AI Fundamentals Learning Guide.
15% Reduction in Context Compression — What You Read Stays Remembered
Previously, as conversations grew longer, AI would summarize and compress earlier content. With 1M token context, this compression is reduced by 15%, according to Anthropic CPO Jon Bell. "AI agents running for hours can now remember what they read at the beginning."
.png)
MRCR v2 benchmark (measuring how accurately AI finds information in long documents): Opus 4.6 scored 78.3%, ranking #1 among all current AI models.
Real-World Use Case
Fintech Company Ramp
Software engineer Anton Biryukov uses Claude Code to simultaneously search databases, monitoring systems, and source code. Previously, compression kicked in at just 100K tokens, causing important details to vanish.
After upgrading to 1M tokens, he can "search without data loss, search again, aggregate edge cases, and propose fixes."
Who Benefits from Claude 1M Token Context?
1. Office Workers with Heavy Document Workloads
Feed entire multi-hundred page documents into AI and ask "summarize," "extract key terms," or "what changed from the previous version?" Learn specific use cases in the AI automation guide.
2. Developers — Claude Code for Large Projects
The problem of forgetting earlier code while navigating files is significantly reduced. Especially effective for refactoring and bug tracking in large projects.
3. Research & Analysis
Feed multiple competitor reports simultaneously for comparative analysis, or read several academic papers at once to identify commonalities and differences.
Available on Azure and Google Cloud Vertex AI
Available on Microsoft Azure Foundry and Google Cloud Vertex AI as well. Developers no longer need beta headers — use 1M tokens with no code changes.
Google Gemini vs Claude — 1M Token Accuracy Comparison
Google's Gemini introduced 1M token context first, but Anthropic emphasizes its lead in accuracy. On the MRCR v2 benchmark, Opus 4.6 scored 78.3% — the highest among all current AI models.
The Future of AI Context Windows
A year ago, 200K tokens was the maximum; now 1M is becoming standard. However, as Anthropic notes, bigger isn't always better due to "context rot" — carefully selecting relevant information remains important.
To learn more about AI work automation, check out our free learning guide.
Related Content — Get Started with AI | Free Learning Guide
Sources
Stay updated on AI news
Simple explanations of the latest AI developments