Amazon Set to Take on OpenAI, Google, and Anthropic with Its Own Reasoning AI Model

Amazon is developing a new artificial intelligence (AI) model with advanced reasoning capabilities. The model is being developed under the Nova brand and could pose serious competition to the main players in the market – OpenAI, Anthropic and Gemini.

Image source: Christian Wiediger / Unsplash

Nova will use a hybrid approach to reasoning, combining fast responses with more complex, multi-tasking thinking in a single system, Business Insider reported, citing a source familiar with the project. One of Amazon’s key goals is to lower the cost of running the model compared to competitors like OpenAI o1, Anthropic Claude 3.7 Sonnet, and Google Gemini 2.0 Flash Thinking. The company has previously said that its current, non-reasoning Nova models are 75% cheaper than third-party offerings available through its Bedrock AI platform.

Nova is being developed by the AGI team, led by chief scientist Rohit Prasad, and to get the model into the top 5 for performance, Amazon tests it on external benchmarks that assess programming and math skills, including SWE, Berkeley Function Calling Leaderboard, and AIME.

It is worth saying that AI models with a reasoning function are gradually becoming a new stage in the development of artificial intelligence. And although they work slower, they are able to solve more complex problems using the search for solutions and the chain of thoughts method. Similar technologies have already been presented by Google, OpenAI and Anthropic. Also, the Chinese company DeepSeek has attracted attention due to the fact that it has found an even more effective approach.

Nova is expected to increase Amazon’s competition with Anthropic, which recently launched the Claude 3.7 Sonnet, which also uses a hybrid approach. Amazon declined to comment, but the AI ​​reasoning model could be up and running by June.