Issue 2026-10-01 · Industry · 开源 · 模型发布 · 研究
Hugging Face releases Olmo-core 3 open MoE training framework

Hugging Face released Olmo-core 3, an open training framework that scales mixture-of-experts models to the trillion-parameter range while preserving computational efficiency. In a benchmark, the expert pool grew from 8 to 128 and total parameters from 4.6B to 47B, with training throughput falling by less than 5%.
Read original ↗- FairStress: New benchmark evaluates fairness in LLM decision-making
- Opinion: AI Has Crossed a Line and Should Be Regulated Like Banks
- Open-source inside-llm visualizes LLM internals and quantization gaps
- Perplexity releases Decider v1.3 decision models
- Hugging Face tutorial builds robot pick-and-place simulation with SimReady assets