Alibaba svela QwenLong-L1-32B: il primo modello di apprendimento per rinforzo per il ragionamento su testi lunghi, in competizione con Claude-3.7.

Alibaba Releases QwenLong-L1-32B: The First Long Text Reasoning Model Trained with Reinforcement Learning

Alibaba officially launched QwenLong-L1-32B on May 27, 2023. This large language model is specifically designed for long context reasoning, marking a significant breakthrough in AI's ability to handle long texts. The model's performance not only surpasses o3-mini and Qwen3-235B-A22B but also reaches a comparable level to Claude-3.7-Sonnet-Thinking.

Technical Innovation Highlights

The biggest technological breakthrough of QwenLong-L1-32B is that it is the world's first long text contextual reasoning model trained through reinforcement learning. Developed based on the QwenLong-L1 framework, it employs advanced algorithms such as GRPO (Group Relative Policy Optimization) and DAPO (Direct Alignment Policy Optimization), combined with a hybrid reward function based on rules and models. These innovations significantly enhance the model's accuracy and efficiency in long context reasoning.

In seven long text contextual document question-answering benchmark tests, QwenLong-L1-32B demonstrated outstanding performance, proving its leading capability in handling complex long text tasks.

Complete Solution System

In addition to the model itself, Alibaba has also launched a complete long text reasoning solution. This solution includes four core components:

High-performance QwenLong-L1-32B model
Specially optimized training dataset
Innovative reinforcement learning training methods
Comprehensive performance evaluation system

This complete solution provides developers and researchers with a full chain of tools from model training to performance evaluation, which is expected to accelerate the industrialization process of long text AI applications.

Industry Impact

The release of QwenLong-L1-32B not only showcases Alibaba's strength in AI technology innovation but also sets a new technical benchmark for the entire industry in the field of long text processing. As the application scenarios for large models continue to expand, the ability to reason with long texts will become one of the key indicators of an AI system's intelligence level. The launch of this model is expected to generate significant application value in areas requiring deep understanding of long texts, such as document analysis, legal research, and academic literature processing.