Amazon Developing "Hybrid Reasoning" AI Model To Compete With ChatGPT, Gemini, DeepSeek

Amazon is developing an artificial intelligence model that exhibits exceptional reasoning capabilities. The AI model is anticipated to exhibit "hybrid reasoning" wherein, it would be able to perform both conventional tasks and those requiring complex intellectual functions utilising a blend of intuitive insights and deliberate, methodical thinking, according to a report from Business Insider. The AI is expected to compete against other major AI models such as Gemini, Chat GPT and DeepSeek.

Unlike the large language model that ChatGPT was built on, which uses a large neural network to churn out quick responses, the Amazon model may use "reinforcement learning", a style of machine learning where the focus is on the model using its reasoning to work out the right answers to questions which would reduce it's rate of errors on queries that require in-depth reasoning and logic.

This would also mean that users would not have to use multiple AI models depending on their needs, but have all the functions anticipated in a singular model.

The model will be launched under the company's Nova lineup of AI models, focusing on courting business from enterprise through partnerships with Amazon Bedrock and Azure AI Foundry, instead of being a consumer product.

The AI model will reportedly be built internally, from the ground up. According to the report, Amazon may unveil this product in June.

Despite the focus on competitive pricing, Amazon aims to maintain premium performance as it's vying to secure the lead in AI ranking benchmarks such as AIME, Chatbot Arena, and Berkeley Function Calling Leaderboard which measure the AI's processing and output capabilities.

Amazon had initially backed a similar AI model developed by Anthropic known as Claude 3.7, which was touted as the first "hybrid reasoning" AI model, which Amazon had raised closed to $3.5 billion for. The AI was noted for solving complex coding problems according to Wired.