List of Flash News about MoE model
| Time | Details |
|---|---|
|
2026-08-27 15:30 UTC |
Z.ai: GLM-5.3-Flash MoE Model Debuts on Chinese Chips
Z.ai GLM-5.3-Flash activates 18B of 320B params via MoE, runs on domestic silicon under export curbs, MIT license, 1M context. |
|
2025-02-24 18:24 UTC |
DeepEP: First Open-Source EP Communication Library for MoE Model Training
According to DeepSeek, DeepEP is the first open-source EP communication library designed for MoE (Mixture of Experts) model training and inference. The library features efficient and optimized all-to-all communication capabilities, and offers both intranode and internode support utilizing NVLink and RDMA technologies. This development could enhance model training performance, potentially influencing trading strategies relying on AI model predictions (DeepSeek, 2025). |