Official source · AI-ASSISTED BRIEF
SGLang v0.5.23 adds TPU7x support, low-latency kernels and Paged Experts
Source publication · YYSI publication
SGLang released v0.5.23, adding Google TPU7x support via a PyTorch-native plugin, opt-in NVIDIA megakernel and AMD monokernel low-latency modes, and Paged Experts for running MoE models larger than GPU memory on a single consumer GPU. The release also reports faster DeepSeek-V4.1 prefill and lower first-token latency, plus LoRA serving on quantized MoE models.
Source & context
github.com · Read the original source ↗
This brief uses automated source/date checks and AI-assisted summarization of an official source. Vendor claims are not independent YYSI test results. Check the original announcement before relying on an important detail.