Allen Institute for AI Unveils Olmo-core 3 to Train Trillion-Parameter Mixture-of-Experts Models Efficiently

Covered by 3 sources
Read onSiliconANGLE· 4 min
Olmo-core 3 is an open-source training framework that enhances efficiency and scalability for large mixture-of-experts language models, enabling models with over one trillion parameters while maintaining high throughput and computational efficiency.

Also covered by 2 sources


More in Machine Learning