China introduces upgraded AI foundation model to accelerate scientifi…

China's Academy of Sciences has released ScienceOne Omni, an upgraded AI foundation model designed for scientific discovery across disciplines, trained on 8 million scientific reasoning samples and outperforming GPT-5.5 and Gemini-3.1-Pro on domain-specific benchmarks. The model is already deployed across 50+ Chinese research institutes, 30+ universities, and state enterprises, with an integrated platform pooling 8,000 professional research tools and running on domestic chips. If you're building domain-specific AI for science or engineering, you now have a published benchmark: ScienceOne Omni beats GPT-5.5 on chemistry, physics and astronomy tasks, and it's already live across 80+ institutions. Your narrow model needs to prove it outperforms this, or get acquired by the labs already using it.