Anthropic: new Claude 3 AI outperforms rivals GPT-4, Gemini 1.0

Anthropic, a startup backed by Google and, on Monday, revealed a suite of artificial intelligence models known as Claude 3, the latest salvo in Silicon Valley's near incessant contest to market still-more powerful technology.

According to Anthropic, the most capable in the family, Claude 3 Opus, outperforms rival models GPT-4 from OpenAI and Gemini 1.0 Ultra from Google on various benchmark exams.

"This is the Rolls-Royce of models, at least at this point in time," CEO Dario Amodei said in an interview.

Advances from Anthropic's competitors have already been previewed or are expected, and Reuters was unable to verify the claims independently.

The news, on the heels of myriad rival announcements and Anthropic's Claude 2 release in July, shows how companies are vying to top charts that rank AI performance, at the same time as business customers are still grappling with how to put such technology to use.

Anthropic's President Daniela Amodei said customers would opt for Claude 3 Opus despite a higher price tag "if they had a need for the most cognitively complex tasks," for instance handling a complicated financial analysis accurately.

Anthropic said it would charge $15 for Claude 3 Opus to take in every 1 million pieces of data known as tokens, and at least five times less for its smaller models to handle the same. By contrast, OpenAI charges $10 for each million tokens entered into its GPT-4 Turbo model.

Anthropic also said Claude 3 is its first "multimodal" AI suite. This means, like some other rival models, Anthropic's AI can respond to text queries and to images, for instance analyzing a photo or chart. Popularized by OpenAI's ChatGPT, such "generative" AI creates new content on command.

"While the Claude 3 model family has advanced on key measures of biological knowledge, cyber-related knowledge, and autonomy compared to previous models, it remains at AI Safety Level 2 (ASL-2) per our Responsible Scaling Policy, the company said.

CEO Amodei said Claude 3 would examine, though not produce, images because "we've seen much less enterprise demand for them."

Google last month paused image-generation features for its Gemini chatbot after the service flubbed on diversity, like inaccurately depicting 1940s Nazi soldiers as having various racial backgrounds rather than being white.

Anthropic said its Claude 3 models would be available via Amazon and Google's cloud platforms. It also would sell access to them directly, in 159 countries, Anthropic said.