Unverified50% confidenceEventExact time
OrcaRouter在单张H200上运行了32个并发评估请求,配置包括FP8 KV cache、MTP以及--max-num-seqs 96
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
9/7/2026
First Seen
Valid until: 9/21/2026
Sources
Related Entities
Related Claims
UnverifiedOrnith家族中9B、35B和397B为开放权重模型,31B暂未开放下载60% similarUnverified单卡RTX 4090即可运行GPT-OSS 20B,单卡A100即可运行GPT-OSS 120B60% similarUnverified普通版的27B模型有一个免费托管版本可用,托管在OrcaRouter上59% similarUnverified本地运行70B模型至少需要配备48GB显存的专业GPU,如NVIDIA A6000或双卡RTX 409058% similarUnverified单台DGX H100系统可提供高达32 petaFLOPS的AI算力58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/869809API
curl https://kongchang.com/api/v1/knowledge/claims/869809MCP
get_claim(id=869809)