Nvidia-backed Reflection will reportedly release an open AI model to rival China's best

Reflection logo with the text Rival to China's best, backed by Nvidia

Key points

  • Reflection's first open-weight model is set to arrive soon
  • It's expected to rival China's best open models, not top US ones
  • It rents Nvidia chips through deals with SpaceX and Nebius

Reflection, an AI startup backed by Nvidia (NVDA), is preparing to release its first open-weight model soon, Axios reported on Oct. 4. The model is expected to trail the most advanced US models at first but compete with the top open-weight models from China.

Open-weight models let developers download and modify the model itself, within its license terms. That option isn't available for the closed models behind Claude and ChatGPT. According to the report, Chinese companies make nearly all the leading open-weight models, though other Western companies are expected to release models this month. Reflection declined to comment.

Reflection wants companies to build their own AI

The startup's bigger goal is what it calls an "AI factory." A company or government would combine its own data, Reflection's models and its own computing power to build a cheap, customized AI system. Banks and the Pentagon are among the big AI users wary of Chinese models because of security risks, and Reflection is trying to fill that gap, the report said.

Reflection has already announced plans to put the idea into practice. In March, it signed a memorandum of understanding with South Korea's Shinsegae Group to build a 250-megawatt AI factory in Korea that will run on Reflection's models and Nvidia chips. Ahead of the launch, Reflection has also met with interested parties in Washington and elsewhere to explain the release and how the AI factory will work, according to the report.

Reflection has lined up computing power from SpaceX and Nebius

In June, Reflection agreed to pay SpaceX (SPCX) $150 million a month, after an initial ramp period, from July 1, 2026, through 2029. The deal gives Reflection access to Nvidia GB300 chips at the Colossus 2 data center, and either company can end it with 90 days' notice after the first three months. In July, Nebius (NBIS) agreed to sell Reflection more than $1 billion in computing capacity through 2029. Nebius stock fell the day that deal was announced.

Nvidia has invested $800 million in Reflection. Through the SpaceX deal, Reflection also rents access to Nvidia hardware. Nvidia CEO Jensen Huang has pushed the AI factory idea for years. In July, Nvidia was among the companies that signed a letter urging Washington not to restrict open-weight AI.

Reflection hasn't said when the model will be released or how it performs. We'll update this story when it does.

Frequently asked questions

What is Reflection AI releasing?

Reflection, an Nvidia-backed startup, is preparing to release its first open-weight AI model soon, Axios reported on Oct. 4, 2026. The model is expected to trail the most advanced US models at first but compete with the top Chinese open-weight models.

What is Reflection's AI factory?

It's Reflection's plan for companies and governments to combine their own data, Reflection's open models and their own computing power into a customized AI system. In March 2026, Reflection signed a memorandum of understanding with Korea's Shinsegae Group to build a 250-megawatt AI factory in South Korea.

Where does Reflection get its computing power?

In June 2026, Reflection agreed to pay SpaceX $150 million a month through 2029 for Nvidia GB300 chips at the Colossus 2 data center. In July, Nebius agreed to sell Reflection more than $1 billion in computing capacity through 2029.

More on NBIS and NVDA

Dennis Singleton
Dennis Singleton

Dennis Singleton was born in Australia and later moved to the United States. He has spent years following the markets, but what keeps his attention is how AI is built. He writes about the companies behind the technology, from semiconductor designers and advanced packaging to photonics, memory, networking, and the hardware powering modern AI. His approach starts with filings, earnings, and industry research, then translates the important details into clear, straightforward analysis without unnecessary hype.