This model is considered the first of its kind in China to reach this advanced level, making it a strong competitor to advanced American models such as “GPT-5 Pro” from OpenAI.
The company explained that the model achieved 100% accuracy in both the American Invitational Mathematics Examination (“AIME 2025”) and the Harvard-Massachusetts Mathematics Tournament (“HMMT”), which are among the most difficult global competitions in the fields of algebra, arithmetic, number theory, and probability, according to SCMP and reported by Arabiya Business.
Intuition Labs, a company specializing in artificial intelligence software, indicated that the performance of the models in these competitions is a real test of their ability to think logically and solve problems, adding that the competition between mathematical models has become a major arena for competition between major companies.
The model relies on the “Qwen3-Max” platform, which includes more than a trillion parameters, and was first launched last September.
Alibaba Cloud confirmed that the performance of the new model matches or exceeds other global models such as “Claude Opus 4” from Anthropic, “V3.1” from DeepSeek, “Grok 4” from xAI, in addition to “GPT-5 Pro” from OpenAI.
A real-world cryptocurrency trading experiment also showed the superiority of “Qwen3-Max,” as it achieved a return of 22.3% on an initial investment of $10,000 within two weeks, while the American models recorded losses, with “GPT-5 Pro” being the worst among them with a loss of 62.7%.
The model is currently available to users through the “Qwen Chat” platform or via the Alibaba Cloud API.
Lin Junyang, a researcher in the development team, confirmed that work is still underway to improve the model’s performance, explaining that “the mission is not over yet.”
