SubQ launches Snap, an AI decision model with constant compute per token, offering 100M free tokens to 10,000 early sign-ups.
AI & Agents ·
SubQ has launched Snap, described as a long-context decision model that maintains constant compute per token. The platform is offering 100M free tokens to 10,000 early sign-ups as part of its initial rollout.
Snap's architecture is designed to handle extended context while keeping computational requirements uniform across token processing, distinguishing it from models where compute scales with input length. The free token allocation targets early adopters, a common strategy in crypto and AI products to bootstrap user engagement and gather feedback during launch phases.
Details about Snap's specific capabilities, pricing structure beyond the free tier, and how the constant-compute architecture compares to existing long-context models remain unclear. The mechanics of token allocation across the 10,000 sign-ups and any conditions or timelines for redemption have not been specified.