# Google opens TPU Sync library for faster KV cache moves

> Google's new open‑source C++ library cuts KV cache transfer latency by over 90% on large MoE models.

Oossa · 2026-11-13 · https://oossa.com/en/google-opens-tpu-sync-library-for-faster-kv-cache-moves

Google released TPU Sync, an open‑source C++ library that moves key‑value (KV) caches and model weights between TPU high‑bandwidth memory, host RAM, storage and the network. The company says the library reduces KV transfer latency by more than 90 % on mixture‑of‑experts models ranging from 235 billion to 480 billion parameters.

## The facts

- Google released TPU Sync on 2026-11-13 ("Fri Nov 13 2026 11:32:26 GMT+0100").
- Latency reduction claimed: >90 % for 235B‑480B MoE models.

## Why it matters

Developers can refresh large language model state on TPUs much faster, lowering pause time during inference or fine‑tuning.

## Sources & references

1. [Google Releases TPU Sync: Open-Source KV Cache Transfer for TPUs](https://www.marktechpost.com/2026/11/13/google-tpu-sync-kv-cache-transfer-library/) – MarkTechPost, 2026-11-13

Last updated: 2026-11-13