Amazon is reportedly working on entering the AI “reasoning” model market.

As Business Insider reports, Amazon is developing an AI model with advanced “reasoning” capabilities, similar to those of OpenAI’s o3-mini and DeepSeek’s R1. The model is expected to launch under Amazon’s Nova brand as early as June, following its introduction at the re:Invent developer conference last year.

These so-called reasoning models employ a step-by-step approach to answering queries, which enhances their reliability in fields such as math and science.

According to the report, Amazon plans to implement a “hybrid” reasoning architecture for its new model, similar to Anthropic’s Claude 3.7 Sonnet. If successful, this approach could enable the model to provide both rapid answers and more complex, extended thinking within a single system.

Amazon also aims to make its Nova reasoning model more cost-efficient than its competitors, as claimed by Business Insider. However, this may be a challenging goal, given DeepSeek’s reputation for offering its models at extremely competitive prices.