agentrl/ReSearch-Qwen-7B
TEXT GENERATIONConcurrency Cost:1Model Size:7.6BQuant:FP8Ctx Length:32kPublished:Mar 27, 2025License:mitArchitecture:Transformer0.0K Open Weights Cold

ReSearch-Qwen-7B is a 7.6 billion parameter Qwen2.5-based language model developed by agentrl, fine-tuned using a novel reinforcement learning framework called ReSearch. This model is specifically designed to learn to reason with search operations, integrating search results into its reasoning chain without relying on supervised reasoning step data. It excels at complex question answering and reasoning tasks by dynamically guiding when and how to perform searches based on its internal thought process.

Loading preview...