Amazon reportedly developing its own AI reasoning model

Amazon is said to be developing its own AI model with advanced reasoning capabilities. The upcoming AI model is slated to be released in June this year as part of the Nova series of generative AI models that was introduced by Amazon in December 2024.

Amazon has reportedly adopted a ‘hybrid reasoning’ approach, where the AI model is capable of providing quick answers as well as more complex outputs after extended ‘thinking’.

Amazon-backed Anthropic has also recently announced that it has developed the world’s first hybrid reasoning model called Claude 3.7 Sonnet. Interestingly, Amazon’s latest model could bring it into closer competition with Anthropic. The e-commerce giant has invested around $8 billion in the Claude-maker so far.

Additionally, Amazon is also prioritising making the Nova reasoning model competitive in terms of pricing. The company is reportedly aiming for its model to be ranked in the top 5 on external benchmarks such as SWE, Berkeley Function Calling Leaderboard, and AIME that are designed to test an AI model’s coding and math skills.

If reports are to be believed, Amazon’s AGI division led by AI researcher Rohit Prasad is tasked with building the new AI reasoning model.