Primary sourceCoreWeave
View sourceEvent ·Multi-source
NVIDIA HGX Rubin NVL8 将登陆 CoreWeave 云平台
CoreWeave 宣布 NVIDIA HGX Rubin NVL8 将在本季度上线其云平台。该平台配备 HBM4 显存,可容纳百万 token 级上下文的 KV 缓存,单节点能同时维持多个智能体会话。第六代 NVLink 将张量并行服务保留在节点内部完成。NVIDIA 预计其每兆瓦 token 吞吐量最高可达 HGX B200 的 10 倍。
Current read
CoreWeave 本季度要上 NVIDIA 的 Rubin NVL8 了,HBM4 能装下百万 token 上下文,跑智能体的吞吐是 B200 的 10 倍,做 Agent 服务可以盯一下。
11 sources79° AI heatLast updated 2026/10/21 17:32:54
Evidence
11 sourcesRelated source 1NVIDIA
View sourceRelated source 2rohanpaul_ai
View sourceRelated source 3Tech in Asia
View sourceRelated source 4LMSYS Org (SGLang)
View sourceRelated source 5berryxia
View sourceRelated source 6PyTorch
View sourceRelated source 7arXiv cs.AI
View sourceRelated source 8coderabbitai
View sourceRelated source 9Teortaxes
View sourceRelated source 10Databricks
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.