SGLang Joins Google Cloud and RadixArk to Bring TPU Inference
SGLang announced a partnership with Google Cloud and RadixArk to bring its open-source LLM inference framework to Google Cloud TPUs. SGL-JAX, built on JAX, has reached production-grade quality, enabling developers to run SGLang on the latest TPUs with seamless migration for models like Gemma, Qwen, and DeepSeek. The initiative aims to provide an out-of-the-box, cost-effective inference acceleration solution for frontier LLMs and diffusion models, making TPUs an efficient inference path.
Confirmed
- Partnership: SGLang community collaborates with Google Cloud and RadixArk to optimize LLM inference on TPUs.
- Technical path: SGL-JAX is production-ready, allowing SGLang to run on latest TPUs.
- Model support: Seamless migration for Gemma, Qwen, DeepSeek, and other open-source models.
Unconfirmed
- Native PyTorch support is planned but no specific timeline has been announced.
Why it matters
- Broader hardware ecosystem: This breaks the single-hardware limitation, offering developers an efficient alternative to GPUs, leveraging TPU architecture for improved efficiency and lower costs.
2026-07-30 ~ 2026-08-01 · 6 related posts
Primary sources
- [source] SGLang Partners with Google Cloud to Bring High-Efficiency Inference to TPU — BanghuaZ · 2026-07-30
- SGLang Partners with Google and RadixArk to Deliver Native TPU Inference for LLMs — ying11231 · 2026-07-31
- SGLang Comes to Google TPUs: Enabling Seamless Migration for Major LLMs — BanghuaZ · 2026-07-31
- Google Partners with RadixArk to Bring SGLang to Cloud TPUs — ying11231 · 2026-07-31