Home / Technology / Anthropic Upgrades Claude AI with Smarter Coding & Vision
Anthropic Upgrades Claude AI with Smarter Coding & Vision
16 Apr
Summary
- Claude Opus 4.7 features enhanced coding and vision capabilities.
- Image processing now supports resolutions up to 2,576 pixels.
- New cybersecurity safeguards block high-risk requests automatically.
Anthropic has launched Claude Opus 4.7, an advanced AI model that refines its predecessor's capabilities, particularly in software engineering and vision. The model demonstrates enhanced performance on complex coding tasks, previously requiring close supervision. Its vision capabilities have been expanded, now processing images with resolutions up to 2,576 pixels on the long edge.
To bolster security, Opus 4.7 includes cyber safeguards that automatically detect and prevent prohibited cybersecurity uses. This version has reduced cyber capabilities compared to the more powerful, yet restricted, Claude Mythos Preview. Professionals can access Opus 4.7 for legitimate cybersecurity tasks via Anthropic's Cyber Verification Program.
Opus 4.7 has shown improved scores on benchmarks like finance agent evaluations and GDPval-AA, reflecting better economically valuable knowledge work. Users may need to adjust prompts due to enhanced instruction following. Anthropic also introduced a new 'xhigh' effort level for finer control over reasoning speed and introduced task budgets in public beta for API users.
