AI 导读
突发新闻🚨：Google 已向我们的智能体推理基准提交了其首个 TPUv7 FP8 vLLM 结果，每美元性能比 GB300 NVL72 FP4 还要高出 1.67 倍。🚀 看到 Google 全力推进将 TPU 的能力对外输出到 Google 之外，令人惊叹。👇️ (1/4)🧵
正文
BREAKING NEWS🚨: Google has submitted its FIRST TPUv7 FP8 vLLM results to our Agentic Inference Benchmark and has demonstrated 1.67x better perf per dollar than even GB300 NVL72 FP4. 🚀
It is amazing to see Google go full steam ahead with externalizing the power of TPUs outside of Google.👇️ (1/4)🧵
来源：SemiAnalysis · x.com