https://derekwang2002.github.io/ai-intelligence/en/projects/olmo-core/ Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations. en Mon, 02 Nov 2026 16:56:00 GMT https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261001-01/ ev-20261001-01-published Thu, 01 Oct 2026 16:01:43 GMT New TRIAL <p>Ai2 released Olmo-core 3, a redesigned open MoE training stack that keeps experts resident on GPUs and combines data, tensor, pipeline and expert parallelism. Ai2 ran a 1.2T-parameter configuration with 58.36B active parameters per token across 512 NVIDIA B300 GPUs. The code is fully open, but the published throughput uses random routing and short measurements rather than sustained training.</p><p><strong>Why read: </strong>Olmo-core 3 is a rare open reference stack for trillion-parameter MoE training.</p><p>Recommendation: TRIAL</p><p>Sources: <a href="https://huggingface.co/blog/allenai/olmocore3">huggingface.co · olmocore3</a> · <a href="https://allenai.org/papers/olmocore3">allenai.org · olmocore3</a> · <a href="https://github.com/allenai/olmo-core">allenai/olmo-core · GitHub</a></p>