Reflection AI raised $2 billion at an $8 billion valuation. Seven months ago, the company was valued at $545 million.
Investors include private and strategic participants who support U.S.-based alternatives to foreign technology. The funding will enable the company to expand its research, purchase additional computing systems, and scale operations.
Reflection AI said the capital will be used for recruiting personnel, enhancing computing capacity, and supporting ongoing research programs. The investment shows the demand for domestically developed systems in both academic and commercial sectors.
Founders Bring DeepMind Expertise to Open Research
Reflection AI was founded in March 2024 by Misha Laskin and Ioannis Antonoglou, both formerly with DeepMind. Laskin led reward modeling for DeepMind’s Gemini project, while Antonoglou co-created AlphaGo.
The company calls itself an open alternative to closed U.S. research centers and as a domestic option compared with Chinese systems like DeepSeek. Laskin said, “DeepSeek and Qwen and all these models are our wake-up call because if we don’t do anything about it, then effectively, the global standard of intelligence will be built by someone else. It won’t be built by America.”
Reflection AI shares model weights publicly so researchers can study and use them. At the same time, the company maintains control over full training processes and datasets. Laskin said, “The model weights anyone can use and start tinkering with them. The infrastructure stack, only a select handful of companies can actually use that.”
The company collaborates with businesses and government organizations that apply its research to practical projects. This structure provides both public access to research and operational oversight.
Building the Team and Systems for Large-Scale Research
Reflection AI has approximately 60 employees, including researchers, engineers, and staff managing computing systems. Many team members previously worked at DeepMind and OpenAI.
The company’s computing systems handle tens of trillions of data points. These systems run tasks across text, images, audio, and video, and support autonomous coding projects.
Reflection AI has not released a fully scaled product yet. Current research focuses on text-based models. The lab has shared studies on autonomous coding agents, demonstrating system performance and practical applications.
The company separates research from operations. Researchers work on development, testing, and model evaluation, while engineers manage computing systems and data pipelines. This structure ensures smooth research and system operations.
The lab collaborates with universities, companies, and government groups, facilitating research projects while keeping sensitive systems controlled.
Reflection AI’s computing systems can train models on tens of trillions of tokens. Misha Laskin, CEO, said, “We built something once thought possible only inside the world’s top labs: a large-scale LLM and reinforcement learning platform capable of training massive Mixture-of-Experts (MoEs) models at frontier scale.”
The company supports U.S.-based research. Public access to model weights and structured systems for commercial work enables U.S. organizations to use advanced tools while the company maintains oversight.
Reflection AI continues to expand staff, research, and computing capacity. Its models and systems handle both research and applied work, with the company maintaining operational control over sensitive infrastructure.
The company offers a domestic option for research and commercial work, combining public sharing, organized staffing, and controlled infrastructure to ensure research remains safe and reliable.
