Introducing Olmo-core 3: Open, scalable training infrastructure for large MoEs

AllenAI introduced Olmo-core 3, open infrastructure for training large mixture-of-experts models.

AllenAI announced Olmo-core 3, described as open, scalable training infrastructure for large mixture-of-experts models. The post was published on the Hugging Face blog. No architecture details, benchmark results, model sizes, or download information were included in the available text.

• AllenAI announced Olmo-core 3 as open training infrastructure.
• The stack is aimed at scalable training of large mixture-of-experts models.
• No benchmarks, model sizes, or release details were included in the available text.