NSED Chat

Preview capacity — responses may take a minute while we bring dedicated GPU inference online. The deliberation protocol is already production-grade; latency here reflects shared-tenant inference, not the architecture. Self-hosted and reserved-capacity deployments run at native speed.