Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs
Hugging Face releases OLMo-core 3, open-source scalable training infrastructure for large MoE models
Hugging Face has released OLMo-core 3, an open-source framework for training large Mixture-of-Experts (MoE) language models at scale. This matters because MoE architectures underpin frontier models like GPT-4 and Gemini, and open training infrastructure lowers the barrier for the broader research community to build competitive models. Coming from a Power Player, this signals continued momentum toward democratizing large-scale AI training.