Anthropic has launched Claude Opus 4.6, marking a significant shift in focus from static inference to dynamic orchestration for AI agents. The release introduces several groundbreaking features designed to enhance long-running AI workflows and complex task handling.

Adaptive Reasoning Controls

One of the most significant additions is the introduction of adaptive reasoning controls. Instead of a simple binary toggle for extended thinking, developers now have access to four adjustable "thinking effort" levels:

  • Low: Fast responses for simple queries
  • Medium: Balanced approach for general tasks
  • High: Deep analysis for complex problems
  • Max: Maximum reasoning depth for the most challenging tasks

This granular control allows developers to calibrate reasoning depth based on task complexity, optimizing for both performance and cost efficiency.

1 Million Token Context Window

Claude Opus 4.6 introduces a 1 million token context window, currently available in beta. This allows the model to process approximately 750,000 words in a single request—enough to analyze entire codebases, lengthy legal documents, or extensive research papers in one pass.

Context Window Comparison

The 1M token context represents a massive leap forward. For perspective, this is equivalent to roughly 15-20 average-length novels or several years of email correspondence.

Context Compaction: Solving "Context Rot"

Perhaps the most innovative feature is context compaction, which automatically summarizes earlier portions of conversations to prevent what Anthropic calls "context rot"—the degradation of response quality as conversations grow longer.

On the MRCR v2 benchmark (Multi-turn Response Consistency and Relevance), Claude Opus 4.6 achieved 76% accuracy compared to Sonnet 4.5's 18.5%. This dramatic improvement means the model can maintain coherent, contextually-aware conversations over much longer periods.

Expanded Output Capacity

The maximum output length has doubled to 128,000 tokens, enabling Claude to generate much longer responses in a single turn. This is particularly valuable for:

  • Generating comprehensive documentation
  • Writing long-form content
  • Producing detailed analysis reports
  • Creating complete code implementations

Platform Availability

Claude Opus 4.6 is immediately available across major cloud platforms:

  • AWS Bedrock
  • Google Cloud Vertex AI
  • Microsoft Foundry
  • Direct API access via Anthropic

Pricing and Access

The new model is available to all API users with existing Claude access. Pricing follows a tiered structure based on thinking effort level and context usage, with detailed information available in Anthropic's updated pricing documentation.

Source: InfoQ →