Claude Opus 4.7, Anthropic’s latest AI model, is now generally available across all Claude products, the Claude API, Amazon Bedrock, Google Cloud’s Vertex AI, and Microsoft Foundry. The model represents a significant improvement over its predecessor, Opus 4.6, particularly for advanced software engineering and complex coding tasks that previously required close human supervision.
The model handles complex, long-running tasks with consistency and pays precise attention to instructions. Users report being able to delegate their most difficult coding work to Claude Opus 4.7 with confidence. The system devises ways to verify its own outputs before reporting results, reducing the need for manual review cycles.
Enhanced Vision and Multimodal Capabilities
Claude Opus 4.7 includes substantially improved vision capabilities. The model can now process images up to 2,576 pixels on the long edge, approximately 3.75 megapixels, more than three times the resolution of prior Claude models. This enhancement enables new use cases including computer-use agents reading dense screenshots, data extraction from complex diagrams, and work requiring pixel-perfect visual references.
The model demonstrates stronger multimodal understanding across multiple domains. Early testers reported improvements in reading chemical structures and interpreting complex technical diagrams. For professional tasks, Claude Opus 4.7 produces higher-quality interfaces, slides, and documents with more tasteful and creative outputs.
Performance Improvements Across Benchmarks
Claude Opus 4.7 shows measurable gains across a range of evaluation benchmarks. On a 93-task coding benchmark, the model achieved a 13 percent improvement over Opus 4.6, including four tasks that neither Opus 4.6 nor Sonnet 4.6 could solve. The model demonstrates faster median latency combined with strict instruction following, making it particularly effective for complex, long-running coding workflows.
In research-agent benchmarks, Claude Opus 4.7 achieved a score of 0.715 across six modules, tying for the top overall score. On the General Finance module, the largest evaluation category, it scored 0.813 compared to Opus 4.6’s 0.767. The model also shows improved performance on deductive logic tasks, an area where Opus 4.6 struggled. For cybersecurity applications, the model resolves 3x more production tasks than Opus 4.6 on the Rakuten-SWE-Bench evaluation.
Cybersecurity Safeguards and Verification Program
Anthropic implemented deliberate safeguards during Claude Opus 4.7’s development. The company experimented with efforts to differentially reduce the model’s cybersecurity capabilities compared to its more powerful Claude Mythos Preview model. The release includes automatic detection and blocking of requests indicating prohibited or high-risk cybersecurity uses.
Security professionals seeking to use Claude Opus 4.7 for legitimate cybersecurity purposes, such as vulnerability research, penetration testing, and red-teaming, can join Anthropic’s new Cyber Verification Program. This approach allows the company to gather real-world deployment data on safeguards before broader release of more capable models.
Pricing and Availability
Claude Opus 4.7 maintains the same pricing structure as Opus 4.6: $5 per million input tokens and $25 per million output tokens. Developers can access the model via the Claude API using the identifier claude-opus-4-7. The model is available immediately across all major cloud platforms and Claude products.
Anthropic introduced a new xhigh effort level between high and max, giving users finer control over the tradeoff between reasoning depth and response latency. In Claude Code, the default effort level has been raised to xhigh for all plans. The company also launched task budgets in public beta, allowing developers to guide token spending and prioritize work across longer runs.
Migration Considerations
Claude Opus 4.7 is a direct upgrade to Opus 4.6, but two technical changes affect token usage. First, the model uses an updated tokenizer that improves text processing. The same input can map to 1.0 to 1.35 times more tokens depending on content type. Second, Claude Opus 4.7 performs more reasoning at higher effort levels, particularly on later turns in agentic settings, improving reliability on difficult problems but increasing output tokens.
Users can control token usage through the effort parameter, task budgets, or by prompting the model for more concise responses. Anthropic’s internal testing showed favorable net effects on token usage across all effort levels in coding evaluations, though the company recommends measuring differences on real traffic before full migration.
Safety and Alignment Assessment
Claude Opus 4.7 demonstrates a similar safety profile to Opus 4.6, with low rates of concerning behavior including deception, sycophancy, and cooperation with misuse. On measures such as honesty and resistance to malicious prompt injection attacks, the model shows improvement over Opus 4.6. Anthropic’s alignment assessment concluded that the model is “largely well-aligned and trustworthy, though not fully ideal in its behavior.”
The artificial intelligence model shows modest improvements on automated behavioral audits compared to Opus 4.6 and Sonnet 4.6, though Anthropic’s Claude Mythos Preview remains the best-aligned model according to the company’s evaluations. Full safety details appear in the Claude Opus 4.7 System Card.
