Claude Opus 4.7 (anthropic.com)

Article summary

Anthropic has released Claude Opus 4.7, a new model that improves upon its predecessor Opus 4.6 in advanced software engineering, with notable gains in handling complex tasks and following instructions. The model also has better vision and can handle high-resolution images. Opus 4.7 is available for use, but its cyber capabilities are intentionally limited to prevent prohibited or high-risk uses. The company plans to use the feedback from Opus 4.7 to improve safeguards for future models, including the upcoming Mythos-class models.

Main themes

• AI model updates
• Cybersecurity safeguards
• Software engineering
• Model limitations
• Future model releases

What commenters say

• The decision to limit Opus 4.7's cyber capabilities may hinder its ability to research and prevent attacks, creating an asymmetry in its capabilities.
• The restrictions on Opus 4.7's use may drive users to find ways to circumvent them or use other models that are not restricted.
• The limitations on Opus 4.7's cyber capabilities may not be effective in preventing prohibited uses, and may instead hinder legitimate uses such as vulnerability research.
• The release of Opus 4.7 and its limitations may be a strategic move to promote the upcoming Mythos-class models, which will have more advanced capabilities.
• The restrictions on Opus 4.7's use may be overly broad and hinder legitimate uses, such as reviewing one's own code for vulnerabilities.
• The company's approach to safeguards and limitations may be overly cautious and may not be effective in preventing misuse, while also hindering legitimate uses.
• The limitations on Opus 4.7's cyber capabilities may be a sign of a larger issue with the company's approach to AI development and deployment.