llm-d releases 1.0 for scalable AI inference

llm-d contributors announced the release of llm-d 1.0, a major milestone for the open source distributed inference project now governed by the CNCF.

• The release adds multi-tenant QoS, zero-downtime operations, a stable v1 configuration API, and agent-aware serving for production AI workloads.
• The announcement also cites ecosystem adoption and benchmarks, including IBM Research’s 11.5x lower inference cost use case and a separate demo showing 3,000 concurrent coding agents and 6.6 million output tokens per minute at peak.
• The article includes supporting quotes from leaders at CNCF, Red Hat, IBM, Google Cloud, CoreWeave, NVIDIA, AMD, Cohere, DigitalOcean, JetBrains, Mistral, and Roblox.
• This is a first-time product release announcement, not a retrospective report; it references prior use cases and integrations rather than announcing a priced commercial deal.