Anthropic is announcing the Claude 3 model family, which includes three state-of-the-art models - Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus, offering increasing levels of intelligence.
Claude 3 Opus outperforms peers on common AI evaluation benchmarks and exhibits near-human levels of comprehension on complex tasks.
All Claude 3 models show increased capabilities in analysis, content creation, code generation, and conversing in non-English languages.
Claude 3 Opus achieves the highest levels of intelligence, while Haiku is the fastest and most cost-effective for its category.
The Claude 3 models have improved in providing nuanced, contextual responses rather than unnecessary refusals.
Claude 3 Opus demonstrates improved accuracy on challenging open-ended questions compared to previous models.
Anthropic is working to address biases, ensure neutrality, and avoid partisan stances in the Claude 3 models.
While more capable than previous models, Claude 3 remains at AI Safety Level 2 according to Anthropic's Responsible Scaling Policy.
Claude 3 is better at following complex instructions and producing structured outputs like JSON.
Anthropic plans to continue advancing the Claude 3 models and releasing new features to enhance their capabilities.
