Launch

Nvidia-backed Reflection AI prepares first open-weight model launch

What's happening? Reflection AI is preparing to release its first open-weight model soon, Axios reported on 4 October 2026, citing people familiar with the plans. The model is expected to initially trail the most advanced US frontier models but to be competitive with the top Chinese open-weight systems, and capable enough for companies to build their own proprietary, low-cost AI on top of it. Axios sources say other Western open-weight models are due to launch in October 2026 as well. A Reflection spokesperson declined to comment.

The AI factory: Reflection's longer-term goal is an "AI factory" product that lets institutions combine their own proprietary data, Reflection's models and their own compute into a customised, cheaper AI system. Hedge funds and trading firms are cited as early interested users. The company has already begun testing the concept through a sovereign AI factory partnership with ShinsegaeDealroom has a profile for this one. Try Dealroom → in South Korea, and in recent weeks has briefed parties in Washington and elsewhere on the model release.

Compute and backing: Reflection has been locking up capacity, signing large deals with Nebius and SpaceXDealroom has a profile for this one. Try Dealroom → to rent Nvidia AI servers. Nvidia, which led Reflection's $2B round at an $8B valuation in October 2025, has strongly supported the move towards a top open system, in line with CEO Jensen Huang's long-held AI factory vision of pairing hardware with open models so enterprises keep ownership of their data. CEO Misha LaskinDealroom has a profile for this one. Try Dealroom → has said it will take time to reach frontier capabilities: "They're kind of like rocket ships... To build a big rocket ship, it takes time."

Why it matters: The most capable open-weight models are almost all Chinese, and Western banks and government users such as the Pentagon are reluctant to adopt them for security reasons. An American open-weight model running on Nvidia GPUs would offer them a cheaper alternative to Anthropic, OpenAI and Google. Open-weight models already take a large share of usage on multi-model platforms, but still a small share of higher-spending enterprise API usage.

Read more: Axios

More top stories