Pangu Large Models
Pangu Large Models

    PanguLargeModels

    ModelArts StudioAdapted to DeepSeek R1/V3 Full-sized，Try Now
    ModelArts Studio Console Documentation
    Five Pangu Foundation Models Bolstering Basic Capabilities
    Pangu NLP Model
    This is a 100-billion parameter model pre-trained on extensive text data, absorbing information from various domains. It is continuously optimized through ongoing learning from large text datasets.
    Pangu CV Model
    This is a CV foundation model built using extensive image and video data along with Pangu's unique techniques. It can be fine-tuned with minimal data to adapt to specific tasks.
    Pangu Scientific Computing Model
    This model is designed to be a high-precision model for cutting-edge fields of scientific research such as meteorology, biomedicine, and chemical materials, thereby driving scientific research breakthroughs, industrial upgrades, and innovation.
    Pangu Multimodal Model
    This model integrates multimodal information, including language and visual data, to achieve image understanding, image generation, and video generation. The latest world model dynamically generates interactive and navigable digital spaces.
    Pangu Prediction Model
    Built on the triplet transformer architecture, this model was pre-trained using structured tasks to acquire cross-modal general knowledge. It can then be fine-tuned on users' structured data for cross-domain downstream tasks, enabling more accurate and quantitative predictions of transaction trends and status types.
    Pangu NLP Model
    This is a 100-billion parameter model pre-trained on extensive text data, absorbing information from various domains. It is continuously optimized through ongoing learning from large text datasets.
    Pangu CV Model
    This is a CV foundation model built using extensive image and video data along with Pangu's unique techniques. It can be fine-tuned with minimal data to adapt to specific tasks.
    Pangu Scientific Computing Model
    This model is designed to be a high-precision model for cutting-edge fields of scientific research such as meteorology, biomedicine, and chemical materials, thereby driving scientific research breakthroughs, industrial upgrades, and innovation.
    Pangu Multimodal Model
    This model integrates multimodal information, including language and visual data, to achieve image understanding, image generation, and video generation. The latest world model dynamically generates interactive and navigable digital spaces.
    Pangu Prediction Model
    Built on the triplet transformer architecture, this model was pre-trained using structured tasks to acquire cross-modal general knowledge. It can then be fine-tuned on users' structured data for cross-domain downstream tasks, enabling more accurate and quantitative predictions of transaction trends and status types.

    ModelArts Studio

    ModelArts Studio

    ModelArts Studio

    ModelArts Studio provides a unified portal for enterprises to access inference services for Pangu models and third-party models, as well as large model services for data engineering, model development, domain-specific application development, and agent development.
    Console
    • ModelArts Studio Large Model Development Platform

    Application Scenarios

    Application Scenarios

    Bank employees need to search for various information, such as personnel details, regulations, and processes, and write numerous reports, emails, and meeting minutes daily. They urgently require large AI models to enhance their office efficiency.

    Bank employees need to search for various information, such as personnel details, regulations, and processes, and write numerous reports, emails, and meeting minutes daily. They urgently require large AI models to enhance their office efficiency.

    Scenario Description

    The Pangu Multimodal Model is able to automatically review and classify electronic materials, particularly image-based ones, to ensure compliance and accuracy when processing various service requests. This capability significantly reduces the workload of administrative staff, such as those in human resources and social security departments, who previously had to manually sort, summarize, and review data. As a result, it greatly enhances the overall efficiency of the department's operations.

    The Pangu Multimodal Model is able to automatically review and classify electronic materials, particularly image-based ones, to ensure compliance and accuracy when processing various service requests. This capability significantly reduces the workload of administrative staff, such as those in human resources and social security departments, who previously had to manually sort, summarize, and review data. As a result, it greatly enhances the overall efficiency of the department's operations.

    Scenario Description

    The Pangu CV Model builds a comprehensive and multi-dimensional city perception system. Hundreds of thousands of class-1/2/3 cameras are widely deployed in key cities, and tens of thousands of cameras are installed in district- and county-level cities. Sharing, aggregating, and analyzing massive video data is critical to constructing the IoT sensing system for smart city governance.

    The Pangu CV Model builds a comprehensive and multi-dimensional city perception system. Hundreds of thousands of class-1/2/3 cameras are widely deployed in key cities, and tens of thousands of cameras are installed in district- and county-level cities. Sharing, aggregating, and analyzing massive video data is critical to constructing the IoT sensing system for smart city governance.

    Scenario Description

    The Pangu Prediction Model assists in decision-making and achieves refined production by providing the optimal coal blending formula. It identifies the optimal formula that maximizes profits while meeting coke quality standards and considering overall economic benefits.

    The Pangu Prediction Model assists in decision-making and achieves refined production by providing the optimal coal blending formula. It identifies the optimal formula that maximizes profits while meeting coke quality standards and considering overall economic benefits.

    Scenario Description

    Pangu Weather Model can accurately predict detailed meteorological features at a much faster speed than traditional numerical weather prediction (NWP) methods. Pangu weather forecasts include geopotential, humidity, wind speed, temperature, sea-level pressure, and precipitation. This model can be directly applied to a wide range of downstream forecasting scenarios. Both the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Meteorological Center of China Meteorological Administration (NMCCMA) have confirmed the advantages of Pangu-Weather.

    Pangu Weather Model can accurately predict detailed meteorological features at a much faster speed than traditional numerical weather prediction (NWP) methods. Pangu weather forecasts include geopotential, humidity, wind speed, temperature, sea-level pressure, and precipitation. This model can be directly applied to a wide range of downstream forecasting scenarios. Both the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Meteorological Center of China Meteorological Administration (NMCCMA) have confirmed the advantages of Pangu-Weather.

    Scenario Description
    Digital Employees

    Bank employees need to search for various information, such as personnel details, regulations, and processes, and write numerous reports, emails, and meeting minutes daily. They urgently require large AI models to enhance their office efficiency.

    Scenario Description
    Meeting Recap

    Based on their generalization capabilities, Pangu models summarize meetings, extract key points to generate highlights, and generate meeting minutes.

    Email Processing

    Pangu models offer intelligent, efficient email processing, including email sorting, writing, and replies.

    Material Writing

    Pangu models generate survey reports or project planning documents based on outlines, project content, and industry requirements, with support for both Chinese and English translation.

    Certificate Review

    The Pangu Multimodal Model is able to automatically review and classify electronic materials, particularly image-based ones, to ensure compliance and accuracy when processing various service requests. This capability significantly reduces the workload of administrative staff, such as those in human resources and social security departments, who previously had to manually sort, summarize, and review data. As a result, it greatly enhances the overall efficiency of the department's operations.

    Scenario Description
    Document Compliance Review

    Pangu models verify whether a user has uploaded the correct type of ID card, household register, and diploma, and whether the name on the uploaded documents matches the user's name.

    Material Type Review

    Pangu models utilize the system's compliance review module to verify whether different report contents are compliant, such as checking for missing fields or unsigned documents.

    Smart City

    The Pangu CV Model builds a comprehensive and multi-dimensional city perception system. Hundreds of thousands of class-1/2/3 cameras are widely deployed in key cities, and tens of thousands of cameras are installed in district- and county-level cities. Sharing, aggregating, and analyzing massive video data is critical to constructing the IoT sensing system for smart city governance.

    Scenario Description
    City Governance

    Detects large crowds, unlicensed roadside stalls, non-compliant billboards, litter, wrongly parked public bicycles, trash can overflow, open trash cans, and more.

    Transportation

    Detects wrong-way driving, license plates, parking violations, vehicle quantities, traffic flow, vehicle types, and more.

    Water Management

    Detects and monitors waterlogging, water levels, drain outlets, trespassing vessels, floating objects in rivers, locker chamber intrusions, and more.

    Smart Coal Blending

    The Pangu Prediction Model assists in decision-making and achieves refined production by providing the optimal coal blending formula. It identifies the optimal formula that maximizes profits while meeting coke quality standards and considering overall economic benefits.

    Scenario Description
    Coke Composition Quality Prediction

    It involves accurately predicting coke composition and quality during coke production, which is of great significance to production efficiency improvement, cost control, and environmental protection.

    Coal Blending Formula Optimization

    In the process of blast furnace ironmaking, the production cost is reduced by selecting and matching coal of different types, properties, and quality.

    Weather Forecast

    Pangu Weather Model can accurately predict detailed meteorological features at a much faster speed than traditional numerical weather prediction (NWP) methods. Pangu weather forecasts include geopotential, humidity, wind speed, temperature, sea-level pressure, and precipitation. This model can be directly applied to a wide range of downstream forecasting scenarios. Both the European Centre for Medium-Range Weather Forecasts (ECMWF) and the National Meteorological Center of China Meteorological Administration (NMCCMA) have confirmed the advantages of Pangu-Weather.

    Scenario Description
    Precipitation Forecast

    Global medium-range precipitation prediction for the next 1 to 7 days, with a surface resolution of 0.25° (25 x 25 km, between 60°N and 60°S) and a temporal resolution of 1 hour. The accumulated precipitation in millimeters within 6 hours of the target time is provided.

    Typhoon Track Prediction

    Global typhoon track prediction for the next 1 to 7 days, with a surface resolution of 0.25° (25 x 25 km) and a temporal resolution of 1 hour.

    Pollutant Concentration Prediction

    7-day predictions of AQIs such as PM10, PM2.5, NO₂, O₃, SO₂, and CO.

    More Services

    More Services

    ModelArts

    One-stop AI development platform

    Content Moderation

    Automatic detection of content in text and images