Alibaba Cloud has released the latest entry in its growing Qwen family of large language models. The new Qwen with Questions (QwQ) is an open-source competitor to OpenAI’s o1 reasoning model. As with competing large reasoning models (LRMs), QwQ can correct its own mistakes, relying on extra compute cycles during inference to assess its responses, making it well suited for reasoning tasks like math and coding. Described as an “experimental research model,” this preview version of QwQ has 32-billion-parameters and a 32,000-token context, leading to speculation that a more powerful iteration is in the offing.

Although Alibaba did not introduce QwQ with a research paper, in a GitHub blog post the company describes its newest model as curious and philosophical. “Like an eternal student of wisdom, it approaches every problem — be it mathematics, code, or knowledge of our world — with genuine wonder and doubt,” Alibaba writes.

As with its earlier Qwen offerings, Alibaba is making QwQ available under an Apache 2.0 license, meaning it can be applied to commercial ends.

Alibaba claims QwQ beats OpenAI’s o1-preview on mathematical problem-solving tests. “It also outperforms o1-mini on GPQA, a benchmark for scientific reasoning” though “is inferior to o1 on the LiveCodeBench coding benchmarks, but still outperforms other frontier models such as GPT-4o and Claude 3.5 Sonnet,” VentureBeat reports.

That said, the Qwen team is candid about its “limitations,” which can include random switching between languages and recursive reasoning loops. But it is only a preview, and a more powerful version is no doubt in the works.

VentureBeat explains how reasoning models work: “by generating more tokens and reviewing their previous responses, the models are more likely to correct potential mistakes.” While the self fact-checking helps LRMs “avoid some of the pitfalls that normally trip up models,” the downside, TechCrunch explains, is “they often take longer to arrive at solutions.”

Marco-o1, another Alibaba reasoning model, “might contain hints of how QwQ might be working,” VentureBeat says, noting that even though it’s open source, the Chinese company hasn’t released information on how it trained the new model.

VentureBeat heralds the LLM era giving way to the age of LRMs, citing OpenAI’s 01 as spurring interest, and discussing a number of Chinese competitors.