Anthropic has announced the launch of its new AI model, Claude Opus 4.7, which the company claims improves upon previous versions but is “broadly less capable” than the more powerful Claude Mythos Preview. Unlike Mythos, which was rolled out to select companies as part of the Project Glasswing cybersecurity initiative earlier this month, Claude Opus 4.7 enhances software engineering and real-world task performance while implementing safeguards to block high-risk cyber uses. This strategic move aligns with Anthropic’s commitment to responsible AI deployment and follows their experimental techniques aimed at reducing cyber capabilities during training while retaining effectiveness in other areas. The new model integrates across Anthropic’s products, APIs, and popular cloud platforms at the same pricing as its predecessor, Claude Opus 4.6.
Anthropic: Anthropic is an AI company focused on developing safe and reliable large language models in the Claude family, emphasizing alignment and risk mitigation in frontier AI systems. It collaborates with major cloud providers and cybersecurity firms to test advanced models responsibly. In this news, Anthropic announced the release of Claude Opus 4.7, positioning it as a safer, broadly available model compared to the more powerful Claude Mythos Preview.
Claude Opus 4.7: Claude Opus 4.7 is Anthropic’s latest flagship AI model, designed for superior performance in software engineering, instruction following, real-world task completion, and handling high-resolution images with file system-based memory. It incorporates automatic safeguards to detect and block high-risk cybersecurity requests during use. The model was released as an improvement over prior versions but intentionally features reduced cyber capabilities relative to Claude Mythos Preview to prioritize safety.
Project Glasswing: Project Glasswing is Anthropic’s initiative partnering with leading tech companies like AWS, Google, Microsoft, Apple, Cisco, and CrowdStrike to deploy AI for defensive cybersecurity and secure critical software. It facilitates early access and evaluation of next-generation models in controlled environments. The project enabled the rollout of Claude Mythos Preview to select firms earlier this month, providing insights for safeguards tested in Claude Opus 4.7.
Claude Mythos Preview: Claude Mythos Preview is Anthropic’s most advanced frontier AI model to date, excelling in cybersecurity tasks such as identifying zero-day vulnerabilities and agentic coding. Due to its potent capabilities, it is restricted to select partners and not generally available. In the announcement, Anthropic highlighted it as more broadly capable than Claude Opus 4.7, particularly in cyber domains, informing safeguard development for future releases.
`json
{
“Partnerships”: “Project Glasswing unites major tech and security organizations to harness AI for cybersecurity enhancements.”,
“Safety Measures”: “Anthropic applied experimental techniques to reduce cyber capabilities in Claude Opus 4.7 during training while maintaining strengths in software engineering and real-world applications.”,
“Deployment Strategy”: “Claude Opus 4.7 is integrated into Anthropic’s products, API, and cloud platforms from Microsoft, Google, and Amazon at the same pricing as its predecessor.”
}
`