How does Claude Opus 4.6 manage performance and cost for different types of tasks?

While the token pricing for Opus 4.6 remains the same as its predecessor, Anthropic has introduced new 'effort controls'. This allows developers to use a parameter to adjust the model's level of 'thinking' from its default setting (high) to medium, providing more granular control over the trade-off between performance on complex problems and the speed and cost associated with simpler queries.

Claude Opus 4.6 Update.

Anthropic has released Claude Opus 4.6, the latest version of its flagship model, introducing significant improvements in coding, reasoning, and autonomous task performance. The update is notable for its state-of-the-art results on several industry benchmarks and for being the first Opus-class model to feature a 1 million token context window in beta. This positions Opus 4.6 as a direct competitor for complex, long-running professional workflows that require sustained reasoning and tool use, particularly in software development and financial analysis.

According to the announcement, Opus 4.6 achieves the highest score on the Terminal-Bench 2.0 agentic coding evaluation and outperforms OpenAI’s GPT-5.2 by 144 Elo points on GDPval-AA, a benchmark for economically valuable knowledge work. The model is available immediately through Anthropic's API and major cloud platforms, with pricing remaining unchanged at $5 per million input tokens and $25 per million output tokens. New developer-focused features include 'adaptive thinking' to adjust computational effort based on task complexity, 'effort controls' to manage the balance between intelligence and cost, and context 'compaction' for longer-running API tasks.

The launch of Opus 4.6 signals a deepening focus in the AI industry on moving beyond conversational interfaces to create systems capable of handling multi-step professional tasks. By emphasizing its performance in agentic coding, legal analysis (achieving 90.2% on BigLaw Bench), and its ability to operate within large codebases, Anthropic is targeting enterprise use cases where reliability and complex reasoning are critical. The model’s improved performance over long contexts suggests the next competitive front for AI will be defined by practical utility in specialized, high-value domains rather than general knowledge alone.

With Opus 4.6, Anthropic is not just iterating on a large language model; it's building a platform for autonomous work. The emphasis on agentic skills, long-context reliability, and enterprise-specific benchmarks indicates a strategic focus on specialized, dependable AI agents designed to execute complex professional workflows with minimal human supervision.