Reflection, an AI startup backed by Nvidia, is preparing to launch a high-performance open-weight model, Axios reported on Oct. 4 local time.

According to the report, Reflection's first model will be unveiled soon. Early performance is expected to lag cutting-edge U.S. models but be competitive with top Chinese open-weight models.

Axios, citing sources, reported that Reflection's open-weight model has enough performance for companies to build their own low-cost AI systems.

If a U.S.-made open-weight model is paired with Nvidia GPUs, individuals and companies will have cheaper alternatives than Anthropic, OpenAI and Google.

· Open-weight AI companies are also emerging in the United States... whether they can check China is 'uncertain'

Most of the most powerful open-weight models currently are provided by Chinese companies.

Reflection is aiming for an AI factory that supports organisations in combining their own data, Reflection models and their own computing resources to build customised, low-cost AI systems.

Reflection is testing the concept by announcing cooperation with South Korea's Shinsegae Group on a sovereign AI factory, Axios reported.

Reflection has also signed large contracts in recent weeks to lease Nvidia AI servers with Nebius and SpaceX.