The release would give developers and enterprises another option for running and customizing AI systems on their own infrastructure. Unlike closed products such as ChatGPT and Claude, open-weight models can generally be downloaded and adapted for specific applications.
Reflection was founded in 2024 by former Google DeepMind researchers Misha Laskin and Ioannis Antonoglou. Laskin previously worked on reward modeling for Gemini, while Antonoglou was a co-creator of AlphaGo.
The startup has attracted substantial investment ahead of its first open-weight release. Nvidia led an $800 million portion of Reflection’s $2 billion Series B in October 2025, when the company was valued at $8 billion. Other investors in that round included Eric Schmidt, Citigroup, 1789 Capital, Lightspeed and Sequoia Capital.
By April 2026, Laskin said Reflection had raised additional financing at a $25 billion pre-money valuation.
The company has also committed significant resources to computing capacity. Reflection signed a deal worth more than $1 billion with Nebius for access to Nvidia GPUs through 2029. It separately reached an agreement with SpaceX tied to the Colossus 2 data center that calls for payments of $150 million per month beginning July 1, 2026 and could total $6.3 billion if the contract runs its full term.
Those agreements give Reflection access to the infrastructure needed to train and operate larger AI systems. They also align with the company’s broader goal of building what it calls an “AI factory,” where organizations combine Reflection models, private data and dedicated computing infrastructure to create customized AI systems.
Reflection has already started testing that concept through a sovereign AI factory partnership with Shinsegae Group in South Korea.
Its currently available product, Asimov, focuses on helping software developers understand and work with large codebases. The planned open-weight model would move Reflection into the broader foundation-model market for the first time.
The company has also held discussions with interested parties in Washington ahead of the release, according to Axios. Western government agencies and businesses have shown reluctance to rely on Chinese open-weight models for sensitive uses, creating an opportunity for companies attempting to build comparable systems in the U.S. and other Western markets.
Reflection CEO Misha Laskin has said reaching the highest levels of AI capability will take time.
“They're kind of like rocket ships,” Laskin told CNBC. “To build a big rocket ship, it takes time.”
Reflection’s first model is also expected to arrive alongside other open-weight releases from Western developers. For now, however, the company has not published enough technical information to independently compare its model with DeepSeek, Qwen or the leading closed systems.
The upcoming launch will be the first major test of whether Reflection can turn its funding, compute commitments and Nvidia backing into a competitive open-weight model.
This analysis is based on reporting from Startup Fortune.
Image courtesy of Reflection AI.
This article was generated with AI assistance and reviewed for accuracy and quality.
