Red Hat AI 3.4: AgentOps & MaaS for Enterprise AI

Red Hat has announced version 3.4 of its AI platform, introducing new tools for managing autonomous AI agents at scale. This update also brings a new Model-as-a-Service capability. The new Model-as-a-Service, or MaaS, gives developers governed access to curated models through a single interface. It uses standard OpenAI-compatible interfaces and allows administrators to track consumption and enforce policies. Underneath, the platform uses the vLLM inference server and llm-d distributed inference engine for model serving. New request prioritization and speculative decoding support aim to improve response speeds. What's interesting is Red Hat AI 3.4 also introduces AgentOps, a set of tools for managing AI agents from development through production. This includes integrated tracing, observability, and cryptographic identity management. The goal is to provide visibility into agent decision-making, addressing security and governance risks as agents operate more independently. Red Hat AI 3.4 is expected to be available later this month. The bottom line is these new features aim to help enterprises move from AI experimentation to production-grade deployment with greater control and operational assurance.