Silicon Valley has a long tradition of raising the money first and figuring out the product later. Reflection AI is taking that to a new level: billions in compute commitments, a $25 billion pre-money valuation, and, according to new reports, a first open-weight model almost ready to ship.
America's Answer to DeepSeek, Allegedly
Axios reported on October 4 that Nvidia-backed Reflection AI is preparing to release its first open-weight model, aimed directly at the Chinese open models that currently dominate the leaderboards: DeepSeek and Alibaba's Qwen. The company was founded in March 2024 by former Google DeepMind researchers Misha Laskin, who led reward modeling on Gemini, and Ioannis Antonoglou, a co-creator of AlphaGo.
The money is substantial. Reflection raised a $2 billion Series B in October 2025 led by Nvidia at an $8 billion valuation, then raised again in April 2026 at a $25 billion pre-money valuation. It has more than $7 billion in compute commitments, including over $1 billion with Nebius and a SpaceX deal worth $150 million a month starting July 1, 2026, which could reach $6.3 billion over its term.
Open Weights, Big Caveats
Here's the honest part: there's no model name, release date, license, model card, or independent benchmark yet. Reports expect the model to be competitive with the top Chinese open models while still trailing the closed frontier systems from OpenAI, Anthropic, and Google. Reflection's only shipped product so far is Asimov, a code-comprehension agent.
The angle most people miss is the business model. Reflection is leaning into Nvidia's "AI factory" idea, where companies buy the chips and own a customizable open-weight model trained on their own data, instead of renting someone else's API. For businesses concerned about data residency, vendor lock-in, or unpredictable per-token pricing, a strong US-built open model is an option they haven't really had before.
Seven billion dollars buys a lot of GPUs. Soon we'll see whether it also buys a spot on the leaderboard.
If the open-weight-versus-API question is starting to come up in your planning meetings, we can help you compare the trade-offs before you commit either way.
Source: Startup Fortune