Anthropic has launched its latest flagship AI model, Claude Opus 4.7, now generally available across its products and API. This iteration represents a substantial leap over its predecessor, Opus 4.6, particularly in advanced software engineering tasks.
Users report a newfound confidence in delegating their most challenging coding work to Opus 4.7, tasks previously requiring extensive human oversight. The model demonstrates rigor and consistency on complex, long-running operations, adhering precisely to instructions and implementing self-verification protocols.
Multimodal capabilities have also seen a significant upgrade, with Opus 4.7 processing images at a higher resolution. Its creative output for professional tasks, including interface design, slide decks, and documents, is notably improved.
Cybersecurity and Safety Measures
While not as broadly capable as the bleeding-edge Claude Mythos Preview, Opus 4.7 serves as a testing ground for new AI model safety and security protocols. Anthropic is deliberately limiting the release of its most powerful models, like the Claude Mythos Preview, and testing safeguards on less advanced versions first.
Opus 4.7's cybersecurity functions are intentionally less advanced than Mythos Preview's. During training, efforts were made to differentially reduce these capabilities. The deployed model includes safeguards that automatically detect and block high-risk cybersecurity use cases.
The daily brief
Get the free daily brief with the day's AI news, new AI startups and funding rounds every morning.
Join 64,000 other subscribers.
Security professionals interested in legitimate applications like vulnerability research can apply to join the new Cyber Verification Program.
Performance and Availability
Early testers have lauded Opus 4.7's ability to catch logical faults early and accelerate execution, a potential game-changer for large-scale financial technology platforms.
The model is available via the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry. Pricing remains consistent with Opus 4.6, at $5 per million input tokens and $25 per million output tokens.
Anthropic highlights Opus 4.7's improved instruction following, noting that prompts tuned for older models may yield different results due to its literal interpretation of commands.
Its enhanced multimodal support allows for higher-resolution image input, up to 2,576 pixels on the long edge, unlocking new use cases for detailed visual analysis.
Opus 4.7 also excels in real-world applications, demonstrating state-of-the-art performance on finance and economics benchmarks, including GDPval-AA.
The model shows improved memory recall across long, multi-session tasks, utilizing file system-based memory to retain context and reduce the need for repeated input.
