Anthropic, a startup backed by Google and Amazon, unveiled a new suite of artificial intelligence models called Claude 3 on Monday, marking the latest development in the ongoing competition among Silicon Valley giants to promote increasingly powerful technology.
According to the startup, the most advanced model in the Claude 3 family, Claude 3 Opus, surpasses rival models such as GPT-4 from OpenAI and Gemini 1.0 Ultra from Google in various benchmark tests.
“This is the Rolls-Royce of models, at least at this point in time,” said CEO Dario Amodei in an interview.
While advances from Anthropic’s competitors have either been previewed or are anticipated, Reuters could not independently verify these claims.
The announcement comes on the heels of numerous competing unveilings and Anthropic’s release of Claude 2 in July. It illustrates the ongoing race among companies to lead in charts ranking AI performance, even as businesses are still figuring out how to effectively leverage such technology.
Anthropic disclosed that it would charge $15 for Claude 3 Opus for processing every 1 million pieces of data, known as tokens. Additionally, its smaller models, offering similar capabilities, would cost at least five times less for handling the same amount of data. In comparison, OpenAI charges $10 for each million tokens processed by its GPT-4 Turbo model.
Claude 3 is also touted as Anthropic’s first “multimodal” AI suite. Like some rival models, it can respond to both text queries and images, analyzing photos or charts. While it won’t generate images, CEO Amodei explained that there is less enterprise demand for them.
Anthropic stated that its Claude 3 models would be accessible through Amazon and Google’s cloud platforms, and it would also directly sell access to them in 159 countries.
