Why QwQ-32B-Preview is the reasoning AI to watch
There’s a new contender in the AI space that’s making waves: QwQ-32B-Preview.
This so-called “reasoning” AI model is being compared to OpenAI o1, and it’s one of the few you can download under a permissive license. For developers and researchers eager to experiment, that’s a significant bonus.
Built by Alibaba’s Qwen team, QwQ-32B-Preview is anything but lightweight. It packs 32.5 billion parameters—think of these as the building blocks of its problem-solving abilities—and can handle prompts of up to 32,000 words; longer than some novels! Tests show it outperforms OpenAI o1-preview and o1-mini on benchmarks like AIME and MATH. For context, AIME uses other AI models to assess performance, while MATH is a collection of word problems.
But the model isn’t just about maths problems or logic puzzles. What sets it apart is how it approaches tasks. QwQ-32B-Preview plans ahead, fact-checks its work, and avoids common AI mistakes. Of course, it’s not flawless—Alibaba acknowledges issues like language switching, occasional loops, and difficulties with “common sense” reasoning. Even so, it represents a step toward more intelligent AI systems.
QwQ-32B-Preview is accessible: You can run or download it via Hugging Face. However, like other Chinese-developed AI, it operates within regulatory boundaries. That means it carefully avoids politically sensitive topics to comply with China’s rules, ensuring it aligns with “core socialist values.”
Alibaba isn’t alone in this space. Meta’s Llama 3.1 is another open-source option, though it takes a different approach by focusing on generative AI rather than reasoning. While both models are innovative, QwQ-32B-Preview specialises in problem-solving with what the company describes as a human-like approach, putting it in the reasoning category.
The competition in AI inside China is intensifying. Companies such as DeepSeek, Shanghai AI Lab, and Kunlun Tech have entered the reasoning AI race, releasing their models at pace. For example, DeepSeek’s r1 claims to outperform OpenAI’s o1 on half of its benchmark tests, particularly in maths and programming. Shanghai AI Lab’s InternThinker takes a structured approach to problem-solving, incorporating steps such as understanding queries, recalling knowledge, planning solutions, and reflecting on its answers.
This surge of activity highlights how quickly Chinese companies are catching up with US tech giants. Xu Liang, an AI entrepreneur from Hangzhou, summed it up: “OpenAI gave the direction; with research, Chinese tech firms are making progress.” The release of QwQ-32B-Preview and its competitors shows how much ground they’re covering.
But this goes beyond just catching up. Reasoning AI marks a change in how models are designed and used. Unlike older AI systems that relied on brute force to generate answers, reasoning models like QwQ-32B-Preview aim to mimic human problem-solving. The approach not only makes them more effective for complex tasks but also expands their potential use cases, like tackling advanced maths or providing detailed financial advice.
Whether it’s solving puzzles, reasoning through intricate problems, or expanding what open-source AI can achieve, one thing is clear: the evolution of AI is accelerating. Buckle up—this is only the beginning.
(Photo by Unsplash)
See also: Alibaba Cloud overhauls AI partner initiative
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.
Explore other upcoming enterprise technology events and webinars powered by TechForge here.
Comments are closed, but trackbacks and pingbacks are open.