The Claude 3 models are also said to be more accurate than previous models and better able to follow multistep instructions. Compared to Claude 2.1, Opus has shown a twofold improvement in accuracy on open-ended questions. The models are also less likely to refuse to answer harmless prompts and have a more nuanced understanding of requests. The company will soon enable citations, making it easier for users to verify answers within reference material. As of Monday, Sonnet is available via claude.ai and Opus is accessible to Claude Pro subscribers, with Haiku to be released soon.
Key takeaways:
- AI startup Anthropic has released the latest version of its Claude family of AI models, Claude 3, which it claims can handle more complicated queries with higher accuracy and enhanced contextual understanding.
- The Claude 3 family includes three models: Claude 3 Haiku, Claude 3 Sonnet and Claude 3 Opus, each with different capabilities and recommended uses. Sonnet is currently available, with Opus accessible to Claude Pro subscribers and Haiku to be released soon.
- Compared to its predecessors, Claude 3 models are more accurate, better able to follow multistep instructions, and less likely to refuse to answer harmless prompts. They also have a 200,000-token context window, with the potential to accept inputs of more than 1 million tokens.
- Anthropic is committed to addressing risks and biases in AI, with Claude 3 showing fewer biases according to the Bias Benchmark for Question Answering. The company also has teams focused on risks including misinformation, child sexual abuse material, and election interference.