YAYI-Ultra
by Wenge ResearchYAYI-Ultra is the flagship enterprise-level large language model developed by Wenge Research. It excels in multi-domain expertise and multimodal content generation, supporting fields such as mathematics, coding, finance, public opinion, traditional Chinese medicine, and security. The model supports inputs of up to 128k tokens, longer context windows, and multimodal capabilities aligned with over 10 million text-image data pairs. It also features multi-turn dialogue role-playing, content security risk control, and the invocation of 10+ intelligent plugins.
What is YAYI-Ultra?
YAYI-Ultra is the flagship enterprise-level large language model developed by Wenge Research. It features robust multi-domain expertise and multimodal content generation capabilities, supporting expert combinations in various fields such as mathematics, coding, finance, public opinion, traditional Chinese medicine, and security.
Main Features of YAYI-Ultra
- Multi-domain Expertise: Supports expert combinations in various fields, alleviating the "seesaw phenomenon" in vertical domain migration.
- Multimodal Content Generation: Supports text and image outputs, extracting relevant information from the knowledge base.
- Long-text Output: Supports inputs of up to 200,000 characters and outputs of up to 100,000 characters.
- Data Analysis and Visualization: Accurately completes data analysis, calculation, and charting tasks.
- Intelligent Planning for Complex Tasks: Enhances the rationality of planning in multi-tool serial invocation scenarios.
- Online Intelligent Creation: Supports online information collection for creation tasks.
Technical Principles of YAYI-Ultra
- Mixed Expert Architecture: Adopts a mixed expert model architecture, providing high-precision, low-energy intelligent solutions for different fields.
- Instruction Fine-tuning and Data-driven Training: Trained on large-scale, high-quality multilingual corpora, combined with instruction fine-tuning technology.
Evaluation Performance of YAYI-Ultra
In the OpenCompass large model public academic ranking, YAYI-Ultra scored 64.5 points, entering the top ten for the first time. Specific performances include:
- Code Generation: Ranked fifth in LiveCodeBench.
- Complex Instruction Understanding: Ranked ninth in IFEval.
- Knowledge Reasoning Ability: Ranked ninth in MMLU-Pro.
Project Addresses of YAYI-Ultra
- Github Repository: https://github.com/wenge-research/YAYI2
- HuggingFace Model Library: https://huggingface.co/wenge-research
Application Scenarios of YAYI-Ultra
- Media Field: Reduces content creation time by 30%-50% and increases content publication frequency by 20%-40%.
- Medical Field: Accurately diagnoses over 500 common diseases with a dialectical reasoning accuracy rate of up to 90%.
- Finance and Taxation Field: Achieves an answer accuracy rate of 90.1%, enabling 24/7 consulting services.
Model Capabilities
Model Type
multimodal
Supported Tasks
Text Generation
Image Generation
Data Analysis
Task Planning
Content Security
Multi-Turn Dialogue
Role-Playing
Intelligent Plugin Invocation
Tags
Large Language Model
Multimodal Content Generation
Enterprise AI
Natural Language Processing
Text Generation
Image Generation
Data Analysis
Task Planning
Content Security
Intelligent Plugins
Usage & Integration
Pricing
free
API Access
Available
License
Open Source
Apache 2.0
Screenshots & Images
Additional Images
Stats
0
Views
0
Likes
3627
GitHub Stars
Invite Gmail Contacts
