Anthropic says its new AI chatbot models outperform OpenAI's GPT-4

With millions in venture funding and support from Google, the AI startup Anthropic unveiled the most recent iteration of its Claude GenAI technology today.

Furthermore, the company claims that the AI chatbot performs better than OpenAI's GPT-4.

Claude 3, the name given to Anthropic's next GenAI, is a family of models that includes Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku, with Opus being the most potent. Anthropic says that all demonstrate "increased capabilities" in analysis and forecasting, along with improved performance on particular benchmarks compared to models like ChatGPT, GPT-4, and Google's Gemini 1.0 Ultra (but not Gemini 1.5 Pro).

Notably, Claude 3 is Anthropic's first multimodal GenAI, which means it can evaluate text in addition to photos, much like certain variants of Gemini and GPT-4. Claude 3 can process images from PDFs, slideshows, and other document types, as well as charts, graphs, and technical diagrams.

Claude 3 is a step ahead of some of its competitors in the GenAI space, as it can evaluate up to 20 photos in a single request. According to Anthropic, this enables it to compare and contrast photos.

However, the image processing of Claude 3 has its limitations.

Because Anthropic is probably concerned about the moral and legal ramifications, it has disabled the models' ability to identify people. Additionally, the business acknowledges that Claude 3 has trouble with tasks requiring spatial reasoning, such as object counting and recognizing an analog clock face, and is prone to errors with "low-quality" images (less than 200 pixels). Claude 3 is unable to provide precise object counts in photos.