According to Business Insider, Amazon is working on an advanced artificial intelligence model that incorporates sophisticated "reasoning" capabilities. This model is expected to be launched under the Nova brand by June. Nova is part of a series of generative AI models introduced by Amazon last year.
This information comes from an insider directly involved in the project, who requested anonymity due to not being authorized to speak with the media. The source revealed that Amazon aims for this new model to use a "hybrid reasoning" approach, enabling it to provide quick responses while handling more complex and in-depth thought processes within the same system. Amazon has declined to comment on the matter.
In recent years, reasoning models have become the next frontier in artificial intelligence. Although these models operate more slowly, they can tackle complex problems by testing multiple solutions and employing chain-of-thought techniques to backtrack. Recently, companies like Google, OpenAI, and Anthropic have released their own reasoning models, with DeepSeek gaining significant attention for its efficient development of similar products.
According to informed sources, one of Amazon's priorities is ensuring that its Nova reasoning model offers a competitive pricing advantage over rivals such as OpenAI's GPT-4, Anthropic's Claude 3.7 Sonnet, and Google's Gemini 2.0 Flash Thinking. Previously, Amazon stated that its existing Nova models are at least 75% cheaper compared to third-party models available through the Bedrock AI development platform.
Additionally, Amazon hopes that its upcoming reasoning model will rank among the top five performers in external benchmarks like the Software Engineer Exam (SWE), Berkeley Function Calling Leaderboard, and AI Math Exam (AIME). These benchmarks evaluate software development and mathematical skills.
This initiative demonstrates Amazon's commitment to investing in its own series of AI models, even as it emphasizes providing multiple model options through Bedrock. Amazon’s AGI team, led by Chief Scientist Rohit Prasad, has been actively developing this new model.
This move also intensifies Amazon’s competition with AI startup Anthropic, which recently launched its latest model, Claude 3.7 Sonnet. This model also uses a hybrid approach, combining rapid responses with extended chain-of-thought outputs.
To date, Amazon has invested $8 billion in Anthropic, and the two companies have been collaborating closely in areas such as AI chips and cloud computing.
