Anthropic released Claude Opus 4.7 just days after unveiling Claude Mythos, a far more powerful cybersecurity-focused model that remains tightly restricted under the Project Glasswing initiative.
Opus 4.7 uses the same core architecture as Mythos but has been intentionally kept less advanced so it cannot be easily used to conduct or automate sophisticated cyberattacks.

The company says internal evaluations still show Claude Mythos Preview as its most capable and best-aligned model overall, particularly for finding and fixing complex software vulnerabilities, which is why access is limited to around 40 partner organisations.
By contrast, Opus 4.7 has gone through full safety hardening and is positioned as the strongest generally available model in Anthropic's Claude family.

Anthropic reports that Opus 4.7 delivers notable improvements in software engineering, especially on difficult, long-running coding tasks that previously required close human supervision.
The model is designed to pay closer attention to instructions, handle complex workflows more consistently, and even devise ways to verify its own output before responding to users.

On the vision side, Claude Opus 4.7 can now process high-resolution images up to 2,576 pixels on the long edge, or roughly 3.75 megapixels, around three times the detail supported by the previous Opus release.
Anthropic says this allows the model to better interpret dense visual information such as charts, screenshots and PDFs for professional and enterprise use.

Claude Opus 4.7 is generally available across all Claude products, via Anthropic's API, and through major cloud platforms including Amazon Bedrock, Google Cloud's Vertex AI and Microsoft's Foundry offering.
Pricing matches the earlier Opus 4.6 model, with input tokens at 5 dollars per million and output tokens at 25 dollars per million, keeping it firmly in the premium tier.

In Anthropic's internal benchmarks, Claude Opus 4.7 is said to outperform OpenAI's GPT-5.4 and Google's Gemini 3.1 Pro on several advanced tasks, while still falling short of the restricted Claude Mythos Preview in overall capability.
The company describes Opus 4.7 as its most capable generally available model for complex reasoning and "agentic" coding, aimed at high-stakes enterprise workloads where reliability matters as much as raw power.

Claude Opus 4.7 arrives at a moment when frontier AI systems are becoming powerful enough to reshape both productivity tools and cybersecurity itself.
By keeping Claude Mythos behind a gate and offering Claude Opus 4.7 as a safer but still ambitious public release, Anthropic is testing whether strict guardrails and strong performance can rise together rather than pull in opposite directions.

If this balance holds, Claude Opus 4.7 could become the everyday workhorse model for companies that want cutting-edge coding and vision capabilities without taking on the full risk of using an unrestricted frontier system.
The next test will be how developers, enterprises and security researchers put this model to work, and whether its safer design proves enough in a rapidly escalating AI race.