Google Open-Sources TPU Sync and XSE Serving Engine for High-Throughput Inference
Google has open-sourced TPU Sync, a C++ library for moving KV cache blocks and model weights between TPU HBM, host DRAM, storage and the network, alongside XSE, the XLA Serving Engine handling six million queries a second across Search, Ads and YouTube Shorts.
Read the full story at InfoQ AI/ML ↗
Timeline · 3 reports
- 2026-11-13 04:16 · DEV Community — AI
Google Open-Sources XSE: 6 Million Queries a Second Without Python in the Path - 2026-11-13 03:24 · DEV Community — AI
Google Just Open-Sourced the Engine Behind Six Million Queries a Second - 2026-11-12 10:06 · InfoQ AI/ML
Google Open-Sources TPU Sync for KV Cache and Weight Transfer on TPUs
More stories
- Google launches Gemini 4 Hydrogen, a faster, lower-cost model for agentic and coding work — Google AI Blog
- Google opens pre-orders for Gemini-powered smart glasses with Warby Parker and Samsung — Google AI Blog
- Google DeepMind unveils AlphaGenome 2 for cell-type-specific DNA variant predictions — Google DeepMind Blog
- Google opens Gemini 4 Carbon to Pro users and API after cybersecurity debut — Google DeepMind Blog
- AI Coding Agents Accelerate Development but Raise Verification and Security Concerns — DEV Community — AI
- Developers build shared tooling to manage multi-agent coding workflows and budgets — DEV Community — AI
- Formal Verification of Signal’s Protocol with Gemini — Google DeepMind Blog
- Advocacy groups sue OpenAI, Google, Anthropic, Meta to freeze frontier AI development — AP News
Get the daily brief of stories like this at 6:30 every morning →