Amazon is reportedly developing a new artificial intelligence (AI) model with advanced reasoning capabilities. According to a report by Business Insider, this AI ‘reasoning’ model is set to be released in June under the ‘Nova’ brand. Notably, the Nova brand is not entirely new. The company introduced the Nova family of generative AI models last year.
And while Alexa recently got AI powered, Amazon is now stepping up its AI efforts to compete directly with Meta AI, OpenAI’s ChatGPT, and Google’s Gemini in the AI race. The report suggests that Amazon’s AI model will be able to understand, analyze, and generate responses more intelligently, working similarly to other competing AI models. It could potentially help in improving problem-solving, decision-making, and contextual understanding. As per the report, Amazon’s AGI division – led by AI researcher Rohit Prasad – is developing this new AI reasoning model.
Speaking of detail, reasoning in AI refers to the ability of models to analyze problems systematically, break them down into logical steps, and evaluate multiple possibilities before concluding. There are already several reasoning AI models on the market, including OpenAI’s O3-mini, DeepSeek R1, and Google’s Gemini 2.0 Flash Thinking.
However, Amazon’s upcoming Nova model will be slightly different from others, as it might come with a ‘hybrid reasoning’ approach designed to offer quick answers as well as complex, extended thinking. Seems like this is Amazon’s latest strategy to enhance its in-house AI models while maintaining a diverse model offering through its Bedrock AI platform.
Interestingly, Amazon’s ‘hybrid’ reasoning AI model sounds similar to Anthropic’s latest released Claude 3.7 Sonnet. It is designed to be both fast and thorough, adapting its responses based on the user’s needs. Notably, Amazon is a prominent investor in the San Francisco-headquartered Anthropic, with a total investment of $8 billion to date.
Talking about the unique aspects of Amazon’s upcoming model, the company wants its Nova reasoning model to be more cost-effective than competing AI models. However, the Chinese AI startup, DeepSeek is already known for offering very cheap AI models. In fact, the DeepSeek-V2 8B model costs $0.0008 per 1,000 tokens—much cheaper than OpenAI’s GPT-4, which costs $0.01–$0.03 per 1,000 tokens.
Additionally, Amazon is developing its AI reasoning model to rank among the top five performers on benchmarks assessing coding, function calling, and mathematical reasoning. In detail, these benchmarks include SWE (Software Engineer benchmark), the Berkeley Function Calling Leaderboard, and AIME (AI Math Evaluations).
Considering the latest trends, the Seattle-headquartered Amazon also does not want to be left behind in the AI race. The company has already announced its plans to invest approximately $100 billion in AI and cloud computing initiatives in 2025.
