genai / news / / VentureBeat
Claude Opus 4.7 exceeds its most direct rivals on key benchmarks.
Anthropic's Claude Opus 4.7 introduces autonomous self-verification, reducing hallucination loops in agentic tasks.
KEY POINTS
- Opus 4.7 leads GPT-5.4 and Gemini 3.1 Pro in knowledge work and agentic coding benchmarks, with an Elo score of 1753.
- The model now supports high-resolution images up to 2,576 pixels, tripling previous visual processing capabilities.
- A new 'effort' parameter and 'task budgets' allow users to control reasoning depth and token spending.
- Opus 4.7's strict instruction adherence requires significant prompt re-tuning, risking errors for legacy applications.
COMPANIES
Summarized by Newsio from VentureBeat. How we summarize →