Anthropic’s Claude Opus 4.7 is here: Check improvements, safety features, and what’s new

Anthropic has officially launched Claude Opus 4.7, the latest iteration of its advanced artificial intelligence model, following the previous version, Opus 4.6. This new model is engineered to assist users with a variety of tasks, including writing, coding, research, and automating workflows. Users can expect significant upgrades in coding efficiency, reasoning capabilities, and safety features. The rollout of Claude Opus 4.7 comes shortly after the introduction of Claude Mythos, a robust cybersecurity-focused model that is currently available in a preview format for select users. In contrast, Claude Opus 4.7 aims to cater to general users, providing a more accessible and reliable AI alternative. Designed to tackle a broad spectrum of both everyday and professional tasks, Claude Opus 4.7 boasts substantial enhancements in areas such as coding and multidisciplinary reasoning. According to Anthropic, this model excels in managing complex, lengthy tasks with precision and consistently follows instructions. It also incorporates mechanisms to verify its outputs before presenting them to users. In addition to its general capabilities, Claude Opus 4.7 is equipped with features that help identify and mitigate risky or harmful requests. The model is designed to prevent misuse related to cybersecurity and is being utilized to evaluate the effectiveness of its safety systems in real-world scenarios. Security professionals can leverage Opus 4.7 for various cybersecurity tasks, including vulnerability research and penetration testing, and are encouraged to participate in Anthropic’s Cyber Verification Program for testing purposes. Performance-wise, Claude Opus 4.7 has demonstrated remarkable advancements. It achieved a score of 64.3% on agentic coding tasks, marking a notable increase from 53.4% in the previous version. Additionally, it attained a score of 87.6% on the verified version of the same benchmark. The model also shows improved capabilities in tool usage and human-computer interaction, achieving scores of 77.3% in scaled tool use and 78% in agentic computer use. However, it still lags behind competitors in specific areas like agentic search. While Claude Opus 4.7's cybersecurity features are slightly lower than those of Opus 4.6, they still fall short compared to the recently released Mythos Preview. Notably, Opus 4.7 enhances instruction adherence, supports high-resolution image generation, and performs better in evaluations related to finance and file system memory. The new Claude Opus 4.7 is now being deployed across all of Anthropic's platforms, including API integrations, Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry.