Claude Opus 4.6 marks a significant upgrade over its predecessor, enhancing coding skills and multitasking abilities. With a 1 million token context window in beta, it can manage larger codebases and perform complex tasks like financial analysis and document creation more effectively. The model excels in several evaluations, including achieving the highest score on Terminal-Bench 2.0 for agentic coding and significantly outperforming OpenAI’s GPT-5.2 on GDPval-AA by 144 Elo points. It also excels at finding hard-to-locate information online, making it a powerful tool for researchers and professionals. Early testers have praised Opus 4.6 for its autonomy and improved reasoning capabilities. It breaks down complex tasks into manageable subtasks, runs tools in parallel, and identifies blockers with precision. The model's ability to navigate large codebases and catch bugs has been noted as "state of the art." Specific feedback highlights its performance in legal reasoning, where it achieved a 90.2% score on the BigLaw Bench evaluation. In cybersecurity investigations, it outperformed Claude 4.5 in 38 out of 40 cases. The new features also include adaptive thinking, which allows the model to adjust its problem-solving approach based on context, and improved controls for developers to manage intelligence, speed, and cost. Users have reported that Opus 4.6 feels less like a tool and more like a collaborative partner, capable of producing polished work with minimal guidance. This model is available through the Claude API and is priced at $5/$25 per million tokens, maintaining accessibility for developers looking to integrate it into their workflows.