Anthropic today released Claude Opus 4.6, its latest artificial intelligence model, boasting significant improvements in coding, autonomous task management, and information retrieval. The upgrade, announced February 5, 2026, introduces a 1 million token context window in beta for Opus-class models, a first for Anthropic.
According to Anthropic, Opus 4.6 demonstrates state-of-the-art performance on several evaluations, including achieving the highest score on the agentic coding evaluation Terminal-Bench 2.0 and leading all other frontier models on Humanity’s Last Exam, a complex multidisciplinary reasoning test. The model similarly outperforms OpenAI’s GPT-5.2 by approximately 144 Elo points on GDPval-AA, an evaluation focused on economically valuable knowledge work in finance, legal, and other sectors. It also surpasses other models on BrowseComp, a benchmark measuring the ability to locate demanding-to-find information online.
The latest model is designed to “plan more carefully, sustain agentic tasks for longer, and work more autonomously,” according to Anthropic, reducing the need for iterative prompting. Improvements extend to code review and debugging, enabling the model to identify and correct its own errors. The release also includes upgrades to Claude’s integration with Microsoft Excel and a research preview of Claude in PowerPoint.
Anthropic emphasized the model’s safety profile, stating that Opus 4.6 exhibits a level of safety comparable to, or exceeding, other leading frontier models. The company’s system card provides detailed information on safety evaluations and rates of misaligned behavior.
Claude Opus 4.6 is currently available on claude.ai, via the Claude API, and across major cloud platforms. Developers can access the model using claude-opus-4-6 through the Claude API. Pricing remains consistent at $5/$25 per million tokens.
Anthropic is offering $50 in free credits to existing Claude users to encourage adoption of the new model.
