Anthropic has unveiled its latest AI model, Claude Sonnet 5, which demonstrates significant advancements over its predecessor, Sonnet 4.6. This new model not only outperforms Sonnet 4.6 across all benchmarks but also surpasses the larger Opus 4.8 in the GDPval-AA v2 knowledge work test, achieving an impressive score of 1,618. In a strategic move, Anthropic has highlighted the performance gap between its models and those currently restricted by the US government, particularly in the realm of cybersecurity tasks, signaling its awareness of the ongoing regulatory discussions in the AI sector. This positioning may serve to underline the competitive edge of Claude Sonnet 5 while also addressing the broader implications of government restrictions on AI advancements.
Source: The Decoder
