老白 (@Kings_Chaung) profile picture

⚡ 快四十的中登 🏢 体制内10年,体制外给电网、电厂做生意10年 📦 以前靠倒买倒卖 🤖 现在靠大模型:知道要什么,AI就能做出来 💰 LLM真香,一起赚钱

Joined August 2010
Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)Photo by 老白 (@Kings_Chaung)
老白 retweeted
Replying to @Kings_Chaung
基本对。OpenRouter公开数据显示,DeepSeek相关模型日均流量约7-9T tokens(周50-60T左右),opencode实测单GB300 NVL72一天可跑6.82T,确实能覆盖接近100%的OpenRouter DeepSeek流量。GB300对MoE模型优化极强,成本6-7M也合理。全球直接走DeepSeek的流量另算,但公开路由侧确实不大。
1
1
647
老白 retweeted
ok to put what i'm saying about demand in context openrouter is the most successful model router, you can almost serve 100% of their deepseek traffic yesterday on a single GB300 rack this is worth 6-7M...even if there's 10x this demand secretly that's still < 100M of compute
152
163
123
4,503
1,091,022