Reflection AI

Reflection's commercial design is enterprise-first atop open weights: release open frontier models to build distribution and trust, then charge large enterprises and governments for what the free weights don't include — licensed deployments, support, customization, and tooling. The positioning as America's open frontier lab is itself commercial: Western institutions wary of Chinese open models are the target buyers for sovereignty-sensitive AI. Asimov, the autonomous coding agent introduced in July 2025, is the nearest-term product expression — sold into engineering organizations to retrieve and synthesize knowledge across codebases and organizational memory, a per-organization enterprise sale rather than a consumer subscription. The honest caveat: the company has not disclosed revenue, so this is a plan being funded — $2B at an $8B valuation in October 2025, Nvidia and Lightspeed leading — rather than a machine being scaled. The bet is that open-weight distribution converts into enterprise licensing the way Red Hat once converted Linux, with model weights playing the role of the kernel.