Alibaba’s Qwen3-Max-Thinking Achieves Perfect Scores in Elite Math Competitions, Signaling a New Era in AI
Alibaba’s artificial intelligence division has unveiled Qwen3-Max-Thinking, an advanced reasoning AI model that recently achieved perfect scores in two of the world’s most challenging mathematics competitions: the American Invitational Mathematics Examination (AIME) and the Harvard-MIT Mathematics Tournament (HMMT). This milestone marks the first time a Chinese-developed AI model has matched or exceeded Western benchmarks in reasoning-intensive academic tests.
### A Leap Forward for China’s AI Ambitions
Built atop Qwen3-Max, Alibaba’s largest AI architecture boasting over one trillion parameters, Qwen3-Max-Thinking represents the company’s boldest effort yet toward creating general-purpose reasoning models. Released in late September, the Qwen3-Max framework is designed to tackle complex problem-solving tasks on a global scale.
For years, competitions like AIME and HMMT have served as unofficial benchmarks for evaluating the reasoning depth and abstract thinking capabilities of large language models (LLMs). Scoring 100% in these contests signals that Qwen3-Max-Thinking is closing the performance gap with top Western-developed AI systems, placing Alibaba’s AI efforts alongside OpenAI’s GPT-5 Pro, which earlier this year also self-reported flawless results in these contests.
### Verification Gaps and Skepticism
Despite the excitement generated by this announcement, significant questions remain about transparency and verification. Alibaba’s claims currently lack third-party confirmation. Neither AIME nor HMMT maintain public leaderboards for AI models, and no independent audits have verified that the results were achieved under closed-book, internet-free conditions — a crucial factor for establishing authenticity.
Experts urge caution, highlighting that without public verification, it is unclear whether Qwen3-Max-Thinking truly achieved perfect accuracy under standardized and uncontaminated conditions. It is also unknown if the AI was tested on the 2025 versions of the contest problems or if it had prior exposure to similar data during training, which could invalidate the results.
Such verification gaps are a known challenge in AI benchmarking, where companies often race to claim superiority in areas like reasoning, coding, and mathematics. Without reproducibility and rigorous controls, the significance of these perfect scores could remain symbolic rather than scientifically robust.
### Opportunities for Developers and Investors
Beyond the benchmarking headlines, Alibaba’s AI strategy has tangible commercial implications. The company has recently opened API access to Qwen3-Max-Thinking, encouraging developers to explore its reasoning capabilities in practical applications.
This development opens new possibilities for software and data teams to optimize cost-performance by dynamically routing workloads between AI providers, based on factors such as pricing, accuracy, and latency. Developers, particularly in the Asia-Pacific region, may find Alibaba’s AI ecosystem attractive as a locally supported alternative to U.S.-based providers, potentially benefiting from competitive pricing and reliable regional infrastructure beyond Singapore.
Investors are watching closely. If Qwen3-Max-Thinking delivers on complex reasoning while remaining affordable, Alibaba could carve out a strong niche among enterprise developers and AI startups seeking alternatives to established Western models. The success of such models may signal a shifting balance in the global AI landscape, where Chinese-developed systems rival or even outperform their Western counterparts in specialized tasks.
—
Alibaba’s announcement of Qwen3-Max-Thinking’s performance underscores the intensifying East-West competition in AI development. While verification challenges remain, the entrance of such high-performing models adds a compelling chapter to the evolving story of global artificial intelligence innovation.
https://coincentral.com/alibabas-qwen-ai-outsmarts-global-peers-in-math-benchmarks/
