Reflection raises $2 billion to be America’s open-borders AI lab, challenges DeepSeek

Reflection, a startup founded last year by former Google DeepMind researchers, has raised $2 billion at a valuation of $8 billion, a massive jump of 15 times its market capitalization. Value: $545 million Just seven months ago. The company, which originally focused on autonomous crypto agents, now positions itself as an open source alternative to closed frontier labs like OpenAI and Anthropic, and a Western equivalent of Chinese AI companies like DeepSeek.

The startup was launched in March 2024 by Misha Laskin, who led the rewards model for DeepMind’s Gemini project, and Ioannis Antonoglou, who co-created AlphaGo, the AI ​​system that beat the world champion in the board game Go in 2016. Their background in developing these highly advanced AI systems is central to their proposition, which is that intelligence talent Proper artificial intelligence can build frontier models beyond established technology giants.

Along with its new round, Reflection announced that it has hired a team of top talent from DeepMind and OpenAI, and built an advanced AI training suite that it promises will be open to everyone. Perhaps most importantly, Reflection says it has “identified a scalable business model that aligns with our open intelligence strategy.”

The Reflection team currently includes about 60 people, most of whom are AI researchers and engineers in the areas of infrastructure, data training and algorithm development, according to Laskin, the company’s CEO. Reflection has secured a computational cluster and hopes to release a parametric language model next year that will be trained on “tens of trillions of tokens,” he told TechCrunch.

“We have built something that was previously thought to be possible only within the best laboratories in the world: a large-scale LLM software and reinforcement learning platform capable of training massive models of mixture of experts (MoEs) at frontier scale,” Reflection said. books In a post on

The Department of Education points to a specific architecture that powers LLMs – systems that previously only large, closed AI laboratories were capable of training at scale. Deep Sick I had great success when I discovered how to train these models at scale and in an open way, followed by Quinn, Kimi and other models in China.

“DeepSeek and Qwen and all these models are our wake-up call because if we don’t do anything about it, the global standard for intelligence will be built by someone else,” Laskin said. “America will not build it.”

Laskin added that this puts the United States and its allies at a disadvantage because companies and sovereign nations will often not use Chinese models due to potential legal ramifications.

“So you can either choose to live at a competitive disadvantage or rise to the occasion,” Laskin said.

American technicians largely celebrated Reflection’s new mission. David Sachs, White House AI official and cryptocurrency czar, Published on X: “It’s great to see more US open source AI models. A significant segment of the global market will prefer the cost, customizability and control that open source offers. We want the US to win this category as well.”

“This is really great news for American open source AI,” Clem Delange, co-founder and CEO of Hugging Face, an open and collaborative platform for AI creators, told TechCrunch about the round. “The challenge now is to demonstrate high agility in sharing open AI models and datasets (similar to what we see from labs that dominate open source AI).”

Reflection’s definition of being “open” seems to focus on access rather than development, similar to Meta with Llama or Mistral strategies. Reflection will release model weights — key parameters that determine how an AI system works — for public use while keeping the full datasets and training pipelines largely proprietary, Laskin said.

“Actually, the most impactful thing is the model weights, because the model weights can be used by anyone and can start modifying them,” Laskin said. “In terms of the infrastructure stack, only a few companies can actually use that.”

This balance is also supported by Reflection’s business model. Researchers will be able to use the models freely, Laskin said, but revenue will come from large organizations that build products on top of Reflection’s models and from governments that develop “sovereign AI” systems, meaning AI models that are developed and controlled by individual countries.

“Once you get into that area where you’re a large organization, by default you want an open model,” Laskin said. “You want something that you’ll have ownership of. You can run it on your own infrastructure. You can control its costs. You can customize it for different workloads. Since you’re paying an ungodly amount of money for AI, you want to be able to make it as optimized as possible, and that’s the market we’re really serving.”

Reflection has not yet released its first model, which will be largely text-based, with multimedia capabilities in the future, according to Laskin. The money from this latest round will be used to acquire the computing resources needed to train the new models, the first of which the company aims to launch early next year.

Investors in Reflection’s latest round include Nvidia, Disruptive, DST, 1789, B Capital, Lightspeed, GIC, Eric Yuan, Eric Schmidt, Citi, Sequoia, CRV, and others.