Anthropic Unveils Claude Opus 4.7 with Coding Leap and Cyber Safeguards

Anthropic has officially launched Claude Opus 4.7, its most capable generally available model to date, with significant improvements in advanced software engineering capabilities and cybersecurity safeguards. The model is now available across all Claude products, the API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry, maintaining the same pricing structure as Opus 4.6 at $5 per million input tokens and $25 per million output tokens.

According to Anthropic's official announcement, Opus 4.7 demonstrates a 13% improvement in coding resolution on a 93-task benchmark compared to Opus 4.6, including four tasks previously unsolvable by either Opus 4.6 or Sonnet 4.6. The model handles complex, long-running coding workflows with greater precision, catching its own logical faults during planning phases and reducing the need for developer supervision on high-stakes tasks. This represents a significant advancement over previous iterations, with Anthropic noting that "Opus 4.7 handles complex, long-running tasks with rigor and consistency, pays precise attention to instructions, and devises ways to verify its own outputs before reporting back."

The model also introduces substantially improved vision capabilities, supporting images at 2576px resolution (up from 1568px in prior versions), with enhanced low-level perception for tasks like image localization and coordinate mapping. This resolution increase enables better artifact understanding in professional workflows, particularly for developers working with screenshots, diagrams, and documentation. As Anthropic explains, "The model's coordinates are 1:1 with actual pixels, so there's no scale-factor math required," simplifying image-based workflows.

Opus 4.7 represents the first model to implement Anthropic's Project Glasswing, a cybersecurity initiative that differentially reduces cyber capabilities compared to their most powerful model, Claude Mythos Preview. During training, Anthropic experimented with efforts to "differentially reduce these capabilities" in Opus 4.7, resulting in a model with "less advanced cyber capabilities" than Mythos Preview. The release includes automatic safeguards that detect and block requests indicating prohibited or high-risk cybersecurity uses, with Anthropic stating that "what we learn from the real-world deployment of these safeguards will help us work towards our eventual goal of a broad release of Mythos-class models."

Security professionals seeking to use Opus 4.7 for legitimate cybersecurity purposes—such as vulnerability research, penetration testing, and red-teaming—are invited to join Anthropic's new Cyber Verification Program. This structured approach addresses concerns about AI models being misused for malicious cyber operations while allowing controlled testing of security applications.

Technical enhancements include a 1 million token context window, adaptive thinking capabilities that adjust processing depth based on task complexity, and a new xhigh effort level for coding and agentic workloads. The model also introduces task budgets (beta), allowing developers to set token allowances for full agentic loops, with Anthropic noting that "the model sees a running countdown and uses it to prioritize work and finish the task gracefully as the budget is consumed."

Anthropic emphasizes that Opus 4.7 is optimized for professional software engineering and complex agentic workflows rather than conversational use. The company states that "Opus 4.7 is the strongest model Hex has evaluated" and that "it correctly reports when data is missing instead of providing plausible-but-incorrect fallbacks," addressing a key limitation in previous versions. This focus on reliability and self-verification represents a strategic shift toward enterprise-grade AI deployment where accuracy matters more than stylistic consistency.

For developers, the model is accessible via the Claude API using the claude-opus-4-7 identifier, with Anthropic confirming that "Opus 4.7 is the first such model" to implement their cybersecurity safeguards. The company's documentation notes that "Claude Opus 4.7 is our most capable generally available model, performing at the frontier across coding, agentic, and knowledge work capabilities," positioning it as the premium option for complex, high-stakes professional tasks.

The release follows Anthropic's strategic approach to model deployment, where more capable models like Mythos Preview remain limited while cybersecurity safeguards are tested on progressively more capable models. This phased rollout strategy reflects the company's commitment to "extensive testing and evaluation" to ensure "the release of Opus 4.7 meets Anthropic's standards for safety, security, and reliability," as detailed in their model card documentation.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn