Open source
Reflection AI
NVIDIA-backed lab founded in 2024 by former Google DeepMind researchers, building open-weight frontier models such as Beam for enterprises and governments.
Reflection AI is an AI lab that builds open-weight models and the software needed to deploy them. It was founded in 2024 by Misha Laskin, who is CEO, and Ioannis Antonoglou, who is CTO. Both previously worked at Google DeepMind on the training and post-training of Gemini models. The company launched publicly in March 2025 with $130 million in seed and Series A funding.
At launch Reflection framed its long-term aim as building Super Intelligence, starting with autonomous coding agents and extending to broader categories of computer work. Its current public positioning is phrased differently: the company says its mission is to “make intelligence open and accessible to all,” and that “open intelligence” means releasing model weights, publishing the science behind its models, and open-sourcing the software used to customise them. It sells models, infrastructure and forward-deployed engineering to enterprises, governments and sovereign entities.
Reflection has become a central part of NVIDIA’s effort to foster an open-model ecosystem outside China, and is a member of NVIDIA’s Nemotron Coalition. Reports in 2026 described it as having raised about $4.6 billion in total at a $25 billion valuation, with NVIDIA, Sequoia and Lightspeed among its investors.
In October 2026 the company announced Beam, its first open-weight model, a sparse mixture-of-experts model aimed at coding and agentic workloads, with weights and a technical report promised under an Apache 2.0 license. Reflection has not claimed to have built a Super Intelligence.
Key facts
- Founded: 2024, by Misha Laskin and Ioannis Antonoglou
- Leadership: Misha Laskin (CEO), Ioannis Antonoglou (CTO)
- Flagship model: Beam (announced October 2026)
- Backers include NVIDIA, Sequoia Capital and Lightspeed
For current details see reflection.ai.