Introducing Claude Opus 4.7

Anthropic has unveiled Claude Opus 4.7. This version is separate from the new model family Claude Mythos Preview, and is instead an update to an existing model.

As an extension of the recently announced Project Glasswing, Anthropic is addressing the risks and benefits of the cybersecurity domain separately. The company stated that it will deploy Mythos Preview on a limited basis and first test new cybersecurity features on lower-capability models, and Opus 4.7 is the first model to receive this testing.

Opus 4.7 includes a security feature that automatically detects and blocks requests suggesting prohibited or high-risk cybersecurity uses. Anthropic explains that lessons learned through real-world deployment will help with the broader future rollout of Mythos-tier models.

Professionals conducting legitimate cybersecurity work, such as:

• Vulnerability research
• Penetration testing
• Red-teaming activities

can participate in the new Cyber Verification Program.

The rollout covers the following:

• All Claude products and the API
• Amazon Bedrock
• Google Cloud Vertex AI
• Microsoft Foundry

Pricing is the same as Opus 4.6, at $5 per 1 million input tokens and $25 per 1 million output tokens. Developers can access it via the Claude API.