Baseten 工程博客：模型性能优化实践汇总
Model performance | Blog
Baseten 工程博客的模型性能栏目汇总了多项推理优化工作，包括让 pyannote 的说话人分离（diarization）模型提速 9.6x、让 Kimi K3 的 tokenization 在百万 token 智能体负载下提速 18x，以及为 DeepSeek V4 Pro 0813 做推理工程优化。该栏目还涵盖托管 RL rollout 的 delta 权重同步优化、Coval 语音 AI 基准测试，以及为 Kimi K3 和 GLM-5.2 构建 API 的实践。
正文
Blog Inference engineering Model performance How Baseten makes pyannote’s diarization models 9.6x faster Matte Lim 1 other Model performance Optimizing delta weight syncs for managed RL rollouts Paras Stefanopoulos Model performance Baseten leads Coval’s voice AI benchmark Ansel Erol 1 other Model performance The efficient frontier of LLM inference Philip Kiely Model performance Agentic kernels in production Brian Li 2 others Model performance Inference engineering for DeepSeek V4 Pro 0813 Model Performance Team Model performance Making Kimi K3 tokenization 18x faster for million-token agentic workloads Michael Feil Model performance How to build a day-0 API for Kimi K3 Model Performance Team Model performance How we built the new fastest API for GLM-5.2 Alex Korte 6 others 1 2 3 ... 7 requestAnimationFrame(function(){$RT=performance.now()}); Popular models