Alibaba has just officially announced its latest reasoning model, and it looks like DeepSeek and OpenAI have something to worry about — if Alibaba’s “hype” is to be believed. The model is open source, so anyone can access it and experiment with it for free.
Called QwQ-32b (Quan-with-Questions), Alibaba's new AI model runs on significantly fewer parameters than almost all major language models today, meaning it requires fewer resources. Alibaba claims QwQ-32b performs at the same level as DeepSeek or OpenAI's o1-mini.
The DeepSeek-R1 large language model (LLM) made waves when it was launched in early February, challenging the gold standard set by ChatGPT, but at a much lower cost. With QwQ-32b, it seems Alibaba wants to push that limit even further.
Technically, DeepSeek-R1 requires 671 billion parameters to run, with 37 billion parameters enabled. Alibaba's QwQ-32b, meanwhile, requires just 32 billion parameters. These numbers may seem abstract to many, but they make a huge difference in computing power: while DeepSeek-R1 requires 1600GB of VRAM to run, QwQ-32b only needs 24GB of VRAM. In most cases, this would require a graphics card like the Nvidia H100 or equivalent, but even the gaming-grade RTX 4090 has 24GB of VRAM. The latest RTX 5090 bumps that up to 32GB.
Some have even suggested that, despite having only 32 billion parameters, QwQ-32B still performs on par with or surpasses R1 in areas such as mathematics, programming, and general problem solving. Alibaba also said that QwQ-32B outperforms OpenAI's o1-mini, which is built with 100 billion parameters. QwQ-32B is available on Hugging Face, the world's largest open-source AI model community.

Alibaba's QwQ-32b is available under the Apache 2.0 license, which means companies and researchers can freely use it. More importantly, we can try it out via Alibaba's Qwen Chat. Like DeepSeek, QwQ-32b has some limitations in terms of topic restrictions. For example, when asked about political issues, Qwen Chat will give a message that this is an inappropriate topic. But there are also some advantages that are immediately obvious.
QwQ-32b seems to give quite detailed answers even to quick and simple questions. This can be good, but in a way it is quite annoying because it provides a lot of unnecessary information that you did not ask for. One plus is that QwQ-32b can show the entire reasoning process, similar to ChatGPT's Deep Thinking feature — but not as deep.
Whether Alibaba's claims will come true remains to be seen, but it looks like ChatGPT and DeepSeek now have a worthy new competitor.