Search
Ant Group Open-Sources Trillion-Parameter Reinforcement Learning High-Performance Weight Exchange Framework Awex
2025-11-20 /
Read about 0 minute
Author：小编
On November 20, 2025, Ant Group announced the open-sourcing of the trillion-parameter reinforcement learning high-performance weight exchange framework Awex. In October this year, Ant Group had already open-sourced two trillion-parameter flagship models: the non-thinking model Ling-1T and the thinking model Ring-1T. Among them, Ring-1T, based on the Awex framework, can complete weight synchronization at the trillion-parameter level within 5-10 seconds on a GPU cluster with thousands of cards.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
