Anthropic Releases Claude Opus 4.7 as It Tests Safer Cyber Guardrails Ahead of Mythos
Opus 4.7 lands across Claude, Bedrock, Vertex AI, and Foundry with unchanged pricing, while Anthropic uses it to trial new cyber safeguards before any broader Mythos rollout.
Overview
Anthropic released Claude Opus 4.7 on April 16, describing it as its latest generally available model and saying it improves advanced software engineering, vision, instruction-following, and professional document work over Opus 4.6, according to Anthropic. The model is available to Claude Pro, Max, Team, and Enterprise users, and through the Claude Platform, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry, with pricing unchanged at $5 per million input tokens and $25 per million output tokens, according to Anthropic and The Economic Times / Reuters.
What We Know
Anthropic also added a new xhigh effort level, task budgets in the Claude Platform API, and a new /ultrareview command for Claude Code, all aimed at making long-running agentic work more controllable, according to Anthropic. Those changes make Opus 4.7 less about a single benchmark jump and more about packaging the model for sustained, multi-step work across coding, analysis, and review workflows, as Anthropic describes the release.
The more important strategic detail is that Opus 4.7 is not Anthropic’s most capable internal model. Anthropic says Claude Mythos Preview is still stronger overall, and that Opus 4.7 is the first model it is using to test new cybersecurity safeguards before a broader Mythos-class rollout, according to Anthropic and The Verge. Anthropic’s earlier Project Glasswing announcement described Mythos Preview as a general-purpose model with unusually strong computer-security capabilities, and said it would be limited to a set of partner organizations to help harden critical software, according to Anthropic and TechCrunch.
This release follows The Machine Herald’s March 29 coverage of the Mythos leak, which first exposed Anthropic’s internal concern that the next generation of Claude-class models could be far more dangerous in offensive cyber contexts.
What We Don’t Know
Anthropic has not said when, or if, Mythos Preview will reach general release; the company has only said that the safeguards tested on Opus 4.7 will inform that future decision, according to Anthropic. It also has not published a full independent assessment of how much safer Opus 4.7 is in practice, so the real test will come from outside users and security researchers rather than launch-day marketing language, an inference from the limited rollout described by The Verge and Anthropic.
Analysis
The release looks like a deliberate split between commercial capability and constrained cyber capability. Opus 4.7 is the product Anthropic can ship broadly today; Mythos is the model it is still treating as too sensitive to open up. That approach may slow the rollout of the company’s strongest systems, but it also gives Anthropic a way to gather real-world evidence before relaxing guardrails on a more capable model.