Qwen Max is a large-scale Mixture-of-Experts (MoE) language model developed by Alibaba Cloud. It excels in language understanding, generation, and task performance across a variety of modalities.
Key Features:
Mixture-of-Experts (MoE) Architecture: Uses 64 specialized "expert" networks, activating only relevant ones per task for efficient processing.
Extensive Multilingual Support: Supports 29 languages, including Chinese, English, and Arabic.
Long-Context Optimization: Supports 32K context windows with 8K generation.
High Stability: Demonstrates high stability in maintaining prompt instructions, with no erroneous replies during extensive testing.
Intended Use:
Qwen Max is designed for a broad range of applications, including:
Complex reasoning tasks.
AI agents and workflows requiring stable instruction following.
Language Support:
Supports 29 languages, including Chinese, English, and Arabic.
Technical Details
Architecture:
64 specialized "expert" networks
Transformer-based architecture.
Training Data:
Over 20 trillion tokens
Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF) methodologies
Diversity and Bias: Fine-tuned using Reinforcement Learning from Human Feedback (RLHF) with over 500,000 human evaluations to improve safety and alignment.
Performance Metrics and Comparison to Other Models:
Usage
Code Samples:
The model is available on the AI/ML API platform as "Qwen Max" .
The Qwen team emphasizes ethical considerations in AI development by promoting transparency regarding the model's capabilities and limitations. The organization encourages responsible usage to prevent misuse or harmful applications of generated content.
Licensing
Qwen Max is available under specific licensing terms provided by Alibaba Cloud. Users are advised to review the licensing information to understand the permissions and restrictions associated with the model's use.