Anthropic eliminates long-context surcharge: 1 million tokens at standard pricing

Anthropic just made a significant move in the AI pricing war: it eliminated the surcharge it charged for using long context windows. Starting March 13, 2026, developers using Claude Opus 4.6 and Sonnet 4.6 pay exactly the same per token, regardless of whether their prompt has 9,000 or 900,000 tokens.

What was the long-context surcharge?

Until this week, if you sent a prompt exceeding 200,000 tokens, Anthropic applied an additional charge that could double the cost per token. It was essentially a tax for using the model's full context window. Many development teams avoided it or absorbed it as an inevitable operational cost.

What exactly changes?

• Opus 4.6: $5 per million input tokens / $25 per million output tokens — regardless of prompt size
• Sonnet 4.6: $3 per million input tokens / $15 per million output tokens — flat rate
• Media limit: The image and PDF page limit per request increased 6x, from 100 to 600

Practical translation: a 900,000-token request is billed at the same per-token rate as a 9,000-token one. Previously, that difference could mean double the cost.

Why does this matter?

There are three use cases where this changes the rules:

1. Advanced RAG and knowledge bases. Retrieval-Augmented Generation systems that inject many documents into context no longer have a price penalizer. You can pass more context without worrying about extra costs.

2. Full codebase analysis. Large repositories that previously required complex chunking strategies to avoid the surcharge can now be passed to the model entirely without economic penalty.

3. Processing legal, financial, or technical documents. Long contracts, audit reports, technical manuals — everything fits in a single API call at the usual price.

The competitive context

This move comes after Google removed similar restrictions on Gemini. Anthropic is responding directly to competitive pressure. The 1-million-token context window already existed technically, but the surcharge made it expensive for intensive use. Now it's accessible at standard pricing.

For teams building products with Claude, this could mean a significant reduction in the monthly API bill, especially in applications where the average prompt size is high.