Reflection debuts Beam, an open-weight AI model

Reflection has released Beam, an open-weight AI model the company positions as a lower-compute-cost rival to Chinese open-weight models, according to TechCrunch.

Per the report, Reflection’s commercial strategy extends beyond Beam itself. The startup is pitching enterprises and sovereign nations on what it calls “AI factories,” a product allowing institutions to build customized, local AI systems by training Reflection’s models on their own proprietary data.

TechCrunch frames the release within a broader competitive context, noting that open-weight models from Chinese developers have gained traction for their accessibility and cost profile. Beam is described as Reflection’s answer to that dynamic, aimed at offering comparable capability at reduced compute expense.

The report identifies the intended customer base as enterprises and sovereign nations, institutions with both the proprietary data and the incentive to run customized, locally controlled AI systems rather than relying on externally hosted models. The “AI factories” framing suggests Reflection is marketing not just a model but a method, a repeatable process for institutions to produce their own tailored systems.

The published account does not specify Beam’s technical benchmarks, parameter count, or release terms beyond its open-weight status. Further details on pricing, availability, and specific enterprise or sovereign partnerships were not included in the cited report.