Reflection says its latest model can out-price China's cheapest AI labs.
Reflection, an AI startup backed by Nvidia, has released an open-weight model called Beam. Beam has 501 billion parameters. Reflection says Beam runs 3-4x cheaper than the best open-weight models coming out of Chinese AI labs, and is positioning the release as a direct challenge to those labs in the open-weight race.
The claim matters because Chinese open-weight models have spent the past year getting attention for being both capable and cheap, pushing Western AI companies to compete on price as well as performance. Nvidia's backing is worth noting too: a chipmaker benefits when more models exist at more price points, because every one of them needs somewhere to run.
A cost figure from the company selling the model isn't an independent benchmark. Treat "3-4x cheaper" as a marketing number until someone outside Reflection checks the math.