Reflection AI Secures $2 Billion in Funding to Launch Open Frontier Lab, Rivals DeepSeek

Reflection AI, a startup founded last year by two former Google DeepMind researchers, has secured $2 billion in funding at an $8 billion valuation. This represents a staggering 15-fold increase from its $545 million valuation just seven months ago. Initially focused on autonomous coding agents, the company is now positioning itself as both an open-source alternative to closed frontier labs like OpenAI and Anthropic, and a Western counterpart to Chinese AI firms such as DeepSeek.

The startup was co-founded in March 2024 by Misha Laskin, who led reward modeling for DeepMind’s Gemini project, and Ioannis Antonoglou, a co-creator of AlphaGo, the AI system that famously defeated the world Go champion in 2016. Their experience in developing advanced AI systems is central to their argument that top AI talent can build cutting-edge models outside of the established tech giants.

Alongside the funding announcement, Reflection AI revealed it has assembled a team of top talent from DeepMind and OpenAI and built an advanced AI training stack it promises to open to the public. Perhaps most significantly, the company states it has "identified a scalable commercial model that aligns with our open intelligence strategy."

According to CEO Misha Laskin, Reflection AI's team currently comprises about 60 people, primarily AI researchers and engineers working on infrastructure, data training, and algorithm development. The company has secured a compute cluster and aims to release a frontier-scale language model next year trained on "tens of trillions of tokens," he told TechCrunch.

"We built something once thought possible only within the world's top labs: a large-scale LLM and reinforcement learning platform capable of training massive Mixture-of-Experts models at a frontier scale," Reflection AI wrote in a post on X. "We witnessed the effectiveness of our approach firsthand when we applied it to autonomous coding. With that milestone achieved, we're now extending these methods to general agentic reasoning."

The Mixture-of-Experts (MoE) architecture powers today's leading LLMs. Until recently, only large, closed AI labs were capable of training these models at scale. DeepSeek had a breakthrough by figuring out how to train them openly at scale, followed by other models like Qwen and Kimi in China.

"Models like DeepSeek and Qwen are our wake-up call. If we don't act, the global standard for intelligence will be set by others, not by America," Laskin said.

Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss the 20th anniversary of TechCrunch, and a chance to learn from the top voices in tech. Grab your ticket before doors open to save up to $444.

Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss a chance to learn from the top voices in tech. Grab your ticket before doors open to save up to $444.

Laskin added that this situation puts the U.S. and its allies at a disadvantage, as enterprises and sovereign states are often reluctant to use Chinese models due to potential legal and compliance risks.

"The choice is clear: accept a competitive disadvantage or rise to the occasion," Laskin stated.

Reflection AI's new mission has been widely celebrated by American technologists. David Sacks, the White House AI and Crypto Czar, posted on X: "It's encouraging to see more American open-source AI models. A significant segment of the global market will prefer the cost, customizability, and control that open source provides. We must ensure the U.S. leads in this category as well."

Clem Delangue, co-founder and CEO of the collaborative AI platform Hugging Face, commented on the funding round to TechCrunch: "This is excellent news for American open-source AI." He added, "The challenge now will be to demonstrate a high velocity of sharing open AI models and datasets, similar to what we see from the labs currently dominating in open-source AI."

Reflection AI's definition of "open" appears to focus on access rather than full collaborative development, a strategy similar to Meta's with Llama or Mistral's. Laskin explained that the company will release model weights—the core parameters that define an AI system's functionality—for public use, while largely keeping its datasets and complete training pipelines proprietary.

"In practice, the model weights are the most impactful component. Anyone can use them and start experimenting," Laskin said. "The infrastructure stack, however, is something only a select few companies can realistically utilize."

This balance also forms the foundation of Reflection AI's business model. Researchers will have free access to the models, but revenue will come from large enterprises building products on top of them and from governments developing "sovereign AI" systems—models developed and controlled by individual nations.

"When you're a large enterprise, you inherently want an open model," Laskin explained. "You want ownership. You want to run it on your own infrastructure, control costs, and customize it for specific workloads. Enterprises investing significant sums in AI want maximum optimization, and that's precisely the market we serve."

Reflection AI has not yet released its first model, which will be primarily text-based with plans for future multimodal capabilities, according to Laskin. The new funding will be used to secure the compute resources necessary for training. The company is targeting an initial model release early next year.

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