Minimax Open-Sources M3.1 Flash AI Model; Google Cloud Launches Agent Compute Engine

Minimax has open-sourced M3.1 Flash, a 428 billion parameter Mixture-of-Experts (MoE) model with a 1 million token context window, boasting 70% lower inference costs and day-one NVIDIA Blackwell support. Concurrently, Google Cloud has launched Agent Compute Engine, providing single-tenant microVMs backed by Titanium silicon for autonomous agents, aiming to reduce resource costs by auto-suspending idle bots and packing more instances per server.