Anthropic Releases Opus 4.7, Not as ‘Broadly Capable’ as Mythos AI

Anthropic has launched Claude Opus 4.7, positioning it as a significant but controlled upgrade. The model shows over 10% improvement in benchmarks for agentic coding—AI that can autonomously plan and execute complex software tasks—and enhanced capabilities in visual reasoning and creative design for presentations. Despite these gains, Anthropic's announcement contained a notable caveat: Opus 4.7 is intentionally not as 'broadly capable' as its forthcoming flagship model, Mythos AI. The company revealed it experimented with ways to 'tone down' Opus 4.7's capabilities before release, particularly in cybersecurity-related areas, marking a shift towards damage control in a high-stakes environment.

This cautious approach is driven by intense pressure. Anthropic is currently managing huge enterprise demand while navigating significant security risks, including concerns that state actors could distill its models for cyberattacks. The more powerful Mythos AI is being shown only to select banking and government institutions in controlled pilots, even as the U.S. government, which recently banned federal use of Claude, reportedly seeks access. This mirrors actions by peers like OpenAI, which has restricted access to its GPT-5.4-Cyber model. For now, Opus 4.7 remains Anthropic's most powerful publicly available model, serving as a bridge that advances enterprise tools like data-to-presentation workflows while the company withholds its peak technology from the open market.