Anthropic has recently launched Claude Opus 4.7, its latest AI model, which is described as highly intelligent but less powerful than the unreleased Claude Mythos. The Opus series, previously leading for Anthropic, now supports advanced coding, multi-step inference, and enhanced visual intelligence.
Opus 4.7 is available through various partners, maintaining the same pricing as its predecessor, 4.6, while using more output tokens due to its higher effort level in processing. Improvements include enhanced capabilities in handling complex coding tasks with greater reliability and reduced errors, such as hallucinations and inaccuracies.
In benchmark testing, Claude Opus 4.7 outperformed other models, aside from Mythos, especially in the Humanity’s Last Test. It scored 46.9%, surpassing many competitors but still falling short of the Mythos score of 56.8%. The model is said to exhibit a safer profile than its predecessor, with reductions in risks associated with incorrect actions and a boost in factual accuracy.
Source link
