QwQ-32B-Preview
QwQ-32B-Preview is an AI model published by 阿里巴巴, released on 2024-11-28, for 推理大模型, with 320.0B parameters, and 32K tokens context length, requiring about 64GB storage, under the Apache 2.0 license.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
QwQ-32B-Preview currently shows benchmark results led by MMLU Pro (79 / 116, score 70.97), AIME 2024 (47 / 62, score 50), MATH-500 (35 / 43, score 90.60). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.
QwQ-32B-Preview 是 Qwen 团队开发的一个实验性研究模型，旨在推进人工智能推理能力。作为一个32B（320亿参数）的预览版模型，它专注于提高复杂问题求解和分析能力。
在多个学术和技术基准测试中，模型展现了显著的性能：
基于其在数学、编程和科学问题解决方面的优势，QwQ-32B-Preview 模型适用于：
该模型代表了人工智能推理能力研究的重要一步，展示了通过深度思考和自我反思提升机器学习模型能力的潜力。
欢迎关注 DataLearner 官方微信，获得最新 AI 技术推送
