Anthropic, the AI startup behind the Claude chatbot, which is backed by Amazon, says that its new Claude 3 AI model sets new industry benchmarks across a wide range of cognitive tasks.
Claude 3 includes three new models, with varying capabilities. Claude 3 Haiku is the least intelligent of the three, Claude 3 Sonnet sits in the middle, while the Claude 3 Opus is the most intelligent model.
According to data obtained by Anthropic, the Claude 3 Opus ranks much higher than OpenAI’s GPT-4 and Google’s Gemini Ultra in several categories like common knowledge, mixed evaluations, coding, reasoning over text and more. In some cases, even the mid-tier Claude 3 Sonnet performed better than GPT-4 and Gemini Ultra. Check out the image below for a breakdown:
Claude 3 Opus even reached “near-human levels” of intelligence and comprehension for some categories, according to Anthropic, before adding that these new models can be used to power live customer chats, auto-completions and data extraction tasks in real-time.
Haiku, the least intelligent of the three new models, is also the fastest and most cost-effective. Sonnet has higher intelligence that Haiku but is slower. However, it is still 2 times faster than the last-gen Claude 2 and Claude 2.1. Opus, the most intelligent of the bunch, performs at speeds similar to Claude 2 and Claude 2.1.
When it comes to processing visual formats, Google’s Gemini Ultra seems to be the clear winner, with Claude 3 models coming in second.
Claude 3 Opus and Sonnet are available to use now in 159 countries, including Canada. “Haiku will be available soon,” says Anthropic.
Read more about the new model here.
Source: Anthropic
