Anthropic is poised to introduce Claude Opus 4.7 along with a novel AI-fueled design tool that enables users to create websites, presentations, and landing pages using straightforward English commands. This announcement reportedly led to a decline in Adobe, Wix, and Figma stock prices on Monday, as reported by The Information.
These innovations might be unveiled within the week, according to an insider familiar with Anthropic’s plans who spoke with The Information. The design tool is aimed at both developers and non-technical users, positioning it against startups such as Gamma and Google’s Stitch.
Anthropic did not reply to Decrypt’s request for a comment.
Interestingly, Opus 4.7 isn’t the most potent model Anthropic has developed. Claude Mythos holds that distinction—a security-centric powerhouse currently being selectively distributed to certain security firms but kept out of public hands.
The UK’s AI Security Institute recently assessed Mythos Preview and discovered its capability to independently conduct complex cyber attacks, surpassing other models in this domain. It became the first AI to finish ‘The Last Ones,’ a 32-step corporate network attack simulation that usually takes human red teams around 20 hours. Mythos accomplished it in three out of ten tries, completing an average of 22 out of 32 steps—compared to Opus 4.6’s 16.
This development is significant beyond enterprise security. The challenge of accurately gauging AI capabilities has become pervasive across the industry. OpenAI recently labeled the top coding benchmark as ‘contaminated,’ yet comparisons persist using those same metrics. A separate ARC-AGI-3 assessment saw Gemini scoring 0.37% and GPT-5.4 achieving 0.26%, while human performance stood at 100%. Consequently, benchmarks are both disputed and still utilized for validation, complicating the interpretation of Opus 4.7’s advancements until Anthropic releases a comprehensive model card.
The connection between Opus and Mythos runs deeper than many realize. Anthropic develops its leading-edge models by refining them based on the Opus series—the same foundational technology that powers public Claude products is rigorously tested and fortified into Mythos. Thus, Opus 4.7 serves as the groundwork that eventually receives cybersecurity enhancements.
Moreover, Anthropic’s initiatives increasingly focus on development and enterprise applications. The release of Claude code through leaks, the introduction of the skills system and MCP protocol, emphasis on agentic AI, and attention to coding benchmarks underscore this trend. Although not officially declared, these developments suggest a shift from being merely an LLM provider towards adopting a full-stack ‘AI studio’ model, where Claude extends beyond text generation to building and deploying entire products.