Anthropic launched its newest widely available AI model, Claude Opus 4.7, on Thursday, April 16, 2026, to provide developers with improved capabilities for complex software engineering and image analysis. The update succeeds version 4.6 by offering better instruction following and creative document generation for enterprise users.
Technical documentation indicates that while Opus 4.7 represents a step forward for the public-facing lineup, it remains less powerful than the company’s restricted Mythos Preview model. Anthropic noted that the new release did not advance its overall performance frontier because Mythos outperformed it across every benchmark evaluation.
Security measures remain a primary focus for the developer, who opted to limit the release of more potent systems to private partners. Anthropic confirmed that corporations like Microsoft, Apple, Google, JPMorgan Chase, and Nvidia currently hold exclusive access to the high-performance Mythos model.
“We stated that we would keep Claude Mythos Preview’s release limited and test new cyber safeguards on less capable models first,” stated Anthropic in a corporate blog post.
The company explained that Opus 4.7 serves as a testing ground for these protective features before a wider rollout of advanced models occurs. This version specifically includes modified cyber capabilities intended to reduce potential misuse compared to the private Mythos version.
“Opus 4.7 is the first such model: its cyber capabilities are not as advanced as those of Mythos Preview (indeed, during its training we experimented with efforts to differentially reduce these capabilities),” Anthropic stated.
To support legitimate research, the organization introduced a new initiative for security specialists. Professionals aiming to utilize AI for vulnerability discovery can apply for the Cyber Verification Program to bypass certain default restrictions.
Early adoption of the model includes companies such as Shopify, Notion, Vercel, and Replit. Anthropic maintained its existing pricing structure for the update, charging $5 per million input tokens and $25 per million output tokens.
Click Here For The Original Source
