Amazon.com Inc (NASDAQ:AMZN) is stepping up its efforts in the artificial intelligence (AI) large language model arms race.
The company is developing a model codenamed 'Olympus' with 2 trillion parameters, according to reporting from Reuters, which would make it one of the largest models ever trained.
By comparison, OpenAI’s GPT-4 model reportedly contains 1 trillion parameters.
Amazon previously trained smaller models, such as one called Titan, and partnered with AI startups including Anthropic and AI21 Labs to make models available to Amazon Web Services (AWS) customers.
A larger model developed in-house could be more attractive to AWS enterprise users, people familiar with the matter said.
The Olympus team is led by Rohit Prasad, Amazon’s previous head of Alexa, who will report directly to CEO Andy Jassy, according to reports. There is not yet a timeline for the launch of Olympus.
Shares of Amazon fell 0.6% Wednesday morning to $141.95.
Contact Andrew Kessel at andrew.kessel@proactiveinvestors.com
Follow him on Twitter @andrew_kessel