Anthropic rolls out Claude Opus 4.7 with gains in coding and reasoning

Anthropic's new AI model improves instruction accuracy, memory usage and multimodal inputs, while introducing safeguards for cybersecurity-related use cases

Anthropic has launched Claude Opus 4.7, the latest version of its Opus model, with a focus on improving software engineering tasks and long-running workflows. The company says the new model builds on Opus 4.6 with better instruction following, more consistent outputs and improved handling of complex coding tasks. The release also introduces updates to multimodal capabilities, memory handling and safety controls.

Claude Opus 4.7: Details

Claude Opus 4.7 is positioned as an upgrade over Opus 4.6, particularly in coding-related tasks. According to the company, the model is better suited for handling complex, multi-step programming work with less supervision, while maintaining consistency over longer sessions. 

 

It also shows improvements in instruction following. Anthropic notes that the model interprets prompts more literally than previous versions, which may require users to adjust existing prompts to avoid unintended results.

  READ: Google integrates Gemini's Personal Intelligence with Nano Banana imaging

  The model includes upgraded vision capabilities, supporting higher-resolution image inputs of up to 2,576 pixels on the long edge. This enables use cases such as analysing detailed screenshots, diagrams and other visual data. 

In benchmark comparisons shared by the company, Opus 4.7 shows improvements over Opus 4.6 across multiple categories, including coding, reasoning and visual tasks, although it remains behind the more advanced Claude Mythos Preview in some areas. 

Anthropic has also introduced changes aimed at improving real-world workflows. The model is better at using file-based memory across sessions, allowing it to retain context and reduce repeated inputs in longer tasks.

Availability

Claude Opus 4.7 is now available across Anthropic’s Claude products and API, as well as on platforms including Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry.