<p>Anthropic has made the 1-million-token context window generally available for Claude Opus 4.6 and Sonnet 4.6, and dropped the long-context pricing surcharge that previously made large prompts significantly more expensive.</p>
<h2>What changed</h2>
<p>Previously, prompts over ~200,000 tokens triggered a premium pricing tier that roughly doubled input costs. That surcharge is gone. Pricing is now flat regardless of prompt length:</p>
<ul>
<li><strong>Opus 4.6:</strong> $5/million input tokens, $25/million output tokens</li>
<li><strong>Sonnet 4.6:</strong> $3/million input tokens, $15/million output tokens</li>
</ul>
<p>A few other changes shipped alongside this:</p>
<ul>
<li>No beta header required for requests over 200K tokens</li>
<li>Full rate limits apply across all context lengths</li>
<li>Up to 600 images or PDF pages per request (up from 100)</li>
<li>Opus 4.6 1M is now the default Opus model for Claude Code users on Max, Team, and Enterprise plans</li>
<li>Claude Code Max, Team, and Enterprise users get 1M context automatically, which means fewer session compactions</li>
</ul>
<p>Pro and Sonnet users can opt in via <code>/extra-usage</code>, and Max/Team/Enterprise users can select the model with <code>/model</code>.</p>
<p>The 1M context window is available on Claude Platform, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry.</p>
<h2>Benchmark claims</h2>
<p>Anthropic says Opus 4.6 scores 78.3% on MRCR v2 and Sonnet 4.6 scores 68.4% on GraphWalks BFS at 1M tokens, which they claim is the highest among frontier models at that context length.</p>
<h2>What this means in practice</h2>
<p>The pricing change has real architectural implications. Retrieval-augmented generation patterns that minimized token usage partly for cost reasons become less necessary. Developers can now send larger code repositories, documents, or datasets in a single prompt without hitting a cost cliff.</p>
<p>That said, at least one developer testing Opus 4.6’s 1M window on long coding sessions found the quality drop-off past 100K tokens still noticeable - worse decisions, worse code, worse instruction-following. The context window is 1M tokens, but the “smart zone” may still be closer to 100K. Worth keeping in mind before rearchitecting everything around stuffing the full window.</p>


Collections

Anthropic has made the 1-million-token context window generally available for Claude Opus 4.6 and Sonnet 4.6, eliminating the previous long-context pricing surcharge. Pricing is now flat: $5/M input tokens for Opus 4.6 and $3/M for Sonnet 4.6, regardless of prompt length. Additional changes include removal of the beta header requirement for requests over 200K tokens, full rate limits across all context lengths, and increased image/PDF limits (up to 600 per request). Opus 4.6 1M is now the default model for Claude Code users on Max, Team, and Enterprise plans. The 1M context is available on Claude Platform, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. Anthropic claims top benchmark scores among frontier models at 1M tokens, though at least one developer reports quality degradation past 100K tokens in practice.

Claude Opus 4.6 and Sonnet 4.6 now support 1M token context windows at standard pricing