Reflection AI is set to launch an open-weight foundation model this month. The startup aims to compete with leading Chinese systems, specifically DeepSeek and Alibaba’s Qwen, which currently dominate open-weight model leaderboards in coding, mathematics, and reasoning.
Founded in March 2024 by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou, the company has secured significant resources. It signed a compute deal with Nebius worth over $1 billion and utilizes capacity from the SpaceX Colossus cluster, including Nvidia server rentals.
The startup previously raised $2 billion at a valuation of $8 billion in October 2025. Subsequent funding rounds have reportedly pushed its valuation to a range between $8 billion and over $25 billion.
Reflection AI’s strategy involves an 'AI factory' model, combining open weights with client data and GPU compute for localized deployments. Its existing product, Asimov, is an agent for code comprehension.
