With 70 billion parameters and 32 billion tokens, evaluated on 10,000 astronomical evaluation datasets, it is currently the most powerful LLM in astronomical knowledge.

Developed by the Huairou Base of the National Astronomical Observatory, it can understand and answer solar physics questions, recognize and analyze solar images.

Highly automated, making observation decisions through AI, reducing manpower consumption in large survey arrays and improving observation efficiency.

A spectral model trained with 2 million high-quality spectrum, achieving joint analysis and mutual prediction between spectra with different resolutions and band coverage.

A basic time-domain variability model trained with a large amount of variability data, which can be used for analysis tasks such as classification, surface gravity estimation, and stellar flare detection.

Based on NADC's massive data resources and data retrieval services, enabling data retrieval and analysis through large language model interaction. More agent services are coming online.