Anthropic mentions Claude Opus 4.6 can 'overthink' on simpler tasks. How can users manage this?

The model defaults to a 'high' effort setting for better results on difficult problems. To address potential overthinking, added cost, or latency on simpler tasks, developers can use the new '/effort' parameter in the API to adjust the model's thinking down to a 'medium' setting, providing more control over the balance between performance, speed, and cost.

Introducing Claude Opus 4.6.

Anthropic today announced the release of Claude Opus 4.6, an upgrade to its flagship model that introduces significant improvements in coding, complex reasoning, and the ability to operate over long contexts. The new model establishes state-of-the-art performance on several key industry benchmarks, notably outperforming OpenAI’s GPT-5.2 on an evaluation of economically valuable work tasks. In a first for its top-tier models, Opus 4.6 also features a 1 million token context window, now available in beta, signaling a focus on handling extensive and complex information.

The update is supported by strong benchmark results, with Opus 4.6 achieving top scores on the agentic coding evaluation Terminal-Bench 2.0 and the multidisciplinary reasoning test Humanity’s Last Exam. For developers, Anthropic has introduced new API controls such as 'adaptive thinking' and an '/effort' parameter to help manage the model's performance against cost and speed, acknowledging that its deeper reasoning can add latency on simpler queries. The model is available immediately on claude.ai and through its API at the same price point as its predecessor, $5/$25 per million input/output tokens.

This release appears to be a deliberate strategic push by Anthropic to position Claude as a platform for autonomous agents capable of executing complex, multi-step professional workflows. The emphasis on sustained agentic tasks, navigating large codebases, and reliable long-context retrieval targets enterprise-level automation. By concentrating on verifiable performance in specialized domains like legal analysis and cybersecurity, Anthropic is intensifying competition around practical, high-value business applications rather than general conversational ability.

Anthropic's strategy with Opus 4.6 is less about chasing general intelligence and more about cornering the market on high-value, long-horizon enterprise tasks. By focusing on agentic coding, complex document analysis, and verifiable performance on professional benchmarks, the company is positioning Claude not as a creative partner, but as a reliable, autonomous digital worker for specialized industries.

AnnouncementsFeb 5, 2026Introducing Claude Opus 4.6We’re upgrading our smartest model. Across agentic coding, computer use, tool use, search, and finance, Opus 4.6 is an industry-leading model, often by wide margin.