AI company Anthropic has officially launched its latest "fully available" model, Claude Opus 4.7. As an upgrade over Opus 4.6, the new model demonstrates superior performance in advanced software engineering tasks, particularly in complex programming scenarios where it reduces reliance on human intervention. Additionally, Anthropic stated that improvements have been made in image understanding, instruction following, and creativity during presentation and document generation. This release follows earlier this month's launch of Claude Mythos Preview by Anthropic. The latter emphasizes cybersecurity capabilities and was hailed by the company as currently possessing the strongest overall performance. However, unlike Mythos Preview, Opus 4.7 does not represent a breakthrough at the frontier of capability boundaries. In its system card, Anthropic explicitly noted that Mythos Preview outperformed Opus 4.7 across all relevant benchmarks, confirming that the former did not advance what the company terms the "capability frontier." For security reasons, Mythos Preview remains accessible exclusively to select partners including Nvidia, JPMorgan Chase, Google, Apple, and Microsoft. According to Anthropic, the company is preparing for broader future deployment of more advanced models by testing new cybersecurity defense mechanisms on lower-capability systems such as Opus 4.7. As reported, Opus 4.7 incorporates enhanced safety measures compared to previous generations, featuring deliberate restrictions specifically targeting cybersecurity-related functionalities. Deployment experiences will inform future widespread releases of Myths-level models. Furthermore, Anthropic introduced a Cybersecurity Validation Program allowing researchers under specific conditions to temporarily relax certain limitations for purposes like vulnerability discovery. In terms of ecosystem adoption, Opus 4.7 has already attracted multiple clients for testing, including Intuit, Harvey, Replit, Cursor, Notion, Shopify, Vercel, and Databricks. Pricing remains unchanged at $5 per million input tokens and $25 per million output tokens.