Amazon is reportedly working on an AI model with advanced reasoning capabilities, set to be released in June as part of its Nova series of generative AI models, which debuted in December 2024.
Unlike traditional AI models that rely solely on chain-of-thought techniques to solve complex problems, Amazon’s approach integrates a “hybrid reasoning” system. This allows the AI to produce quick responses while also engaging in deeper, extended reasoning when required. By doing so, the model is expected to be more efficient in handling diverse problem-solving scenarios.
Amazon is also focusing on ensuring the Nova reasoning model remains cost-competitive without compromising on performance. The company aims to position its model among the top five on key external benchmarks, such as SWE, the Berkeley Function Calling Leaderboard, and AIME, which evaluate AI models’ coding proficiency and mathematical reasoning.
The development of this AI model is being spearheaded by Amazon’s AGI division, led by AI researcher Rohit Prasad. With this initiative, Amazon is striving to strengthen its foothold in the rapidly evolving generative AI landscape.
