书生通用大模型

What is 书生通用大模型?

书生通用大模型 is a suite of open-source foundation models spanning language, multimodal, weather, aerodynamic, 3D spatial, financial, and spectroscopic domains. The vendor claims the InternLM2.5 large language model processes up to 1 million tokens of context and was pretrained on 5 trillion tokens. Access methods for individual components are not specified in the source.

What are the features of 书生通用大模型?

• Large language model (InternLM2.5)

  Supports a 1 million token context window, autonomous web search across over 100 pages, and parameter scales from hundreds of millions to hundreds of billions.
• Multimodal model (InternVL)

  Open-source multimodal models ranging from edge-deployable sizes to 100-billion-parameter versions, with the smallest model fitting on a single consumer-grade GPU.
• Weather and ocean forecasting (FengWu)

  AI models for short-term precipitation, global medium-range forecasts exceeding 10 days at 10-kilometer resolution, and interannual ocean predictions.
• Airfoil generation (YiFei)

  Generates and edits supercritical airfoils using drag-and-drop, text, or voice input, embedding aerodynamic, structural, and material constraints.
• 3D spatial modeling (TianJi)

  NeRF-based system reconstructing city-scale scenes up to 100 square kilometers with 4K training precision and real-time rendering at 1K resolution and 30 frames per second.
• Spectrum analysis (YanPu IS-EPR)

  Parses electron paramagnetic resonance spectra for radicals, multi-species, and metals, linking results to global literature databases.

What are the use cases of 书生通用大模型?

• Financial text analysis

  JiShi applies retrieval-augmented generation and compliance checks to process financial documents using 193 billion tokens of domain-specific pretraining data.
• Civil aircraft design

  YiFei generates supercritical airfoils with a claimed geometric error rate of 1‰ to reduce iteration cycles in complex aerodynamic design.
• Scientific spectrum interpretation

  YanPu reduces EPR spectrum parsing time from hours to seconds for applications in physics, chemistry, materials science, and biology.

书生通用大模型 Contact Email:

intern-ai@pjlab.org.cn