NVIDIA Nemotron 3.5 Lightning 上线 Together AI

The non-obvious point is that RL cadence, not compute size, becomes the serving bottleneck once weights update every few hours, and hot-loading removes the downtime tradeoff. ML platform engineers and inference teams running post-training pipelines should attend to copy the Dynamo and RL Rollouts setup directly, including cost figures. Comparing on-policy weight transition approaches before committing to an orchestration stack could save a costly rework later.

CoreWeave@CoreWeaveSource

Hardware vendors co-funding a single open-source inference engine reduces the risk that serving optimizations remain locked to one chip ecosystem, which matters for teams planning multi-vendor GPU fleets. Engineering leads running large-scale inference can track the consortium's roadmap to anticipate support for new models and distributed serving features. Organizations in Southeast Asia building inference capacity on heterogeneous hardware will benefit from upstream optimizations landing in SGLang rather than vendor-specific forks.

Official accountLMSYS Org (SGLang)@lmsysorgSource

The 200 kW-per-rack threshold marks a shift from air cooling to fully engineered liquid-cooled racks as a prerequisite for hosting NVIDIA's densest platforms. AI infrastructure teams and GPU cloud buyers should study the pump and heat-exchanger sizing approach, since cooling reliability now determines whether a provider can offer Vera Rubin capacity at scale. Teams planning data center builds for NVL72-class systems get a concrete reference for what Mission Control-style rack orchestration must handle.

CoreWeave@CoreWeaveSource