Alibaba launches Qwen-3 Max, its most powerful AI model yet to rival ChatGPT

Alibaba has officially introduced the Qwen-3 Max, marking a significant advancement in its series of large language models. This new iteration is designed to compete head-to-head with formidable players in the AI field, such as OpenAI's GPT-5, Google’s Gemini 2.5 Pro, and Claude’s Opus 4. What sets Qwen-3 Max apart is its unprecedented scale, being the first model in Alibaba's Qwen series to exceed one trillion parameters. This remarkable feat was achieved through pretraining on a staggering 36 trillion tokens, enabling the model to handle context lengths of up to one million tokens. Such capacity allows it to process entire codebases or extensive documents with ease. The advancements in Qwen-3 Max include enhanced reasoning abilities, improved instruction adherence, and robust multilingual support. Additionally, the model is reported to excel in mathematics, coding, logic, and science tasks, with a particular focus on delivering accurate responses in both English and Chinese. Alibaba claims that the new model significantly reduces the occurrence of hallucinations—instances where the AI fabricates information—and provides higher quality responses to open-ended inquiries. In terms of performance rankings, Qwen-3 Max currently stands at third place on LMArena’s text leaderboard, trailing behind Gemini 2.5 Pro and Claude Opus 4.1, but surpassing the standard GPT-5 model. On the SWE-Bench Verified benchmark, which assesses AI's capability to tackle real-world coding challenges, Qwen-3 Max scored 69.6, outperforming the DeepSeek V3.1 non-thinking model but falling slightly short of Claude Opus 4. In the Tau2-Bench evaluation, focusing on agent tool-calling proficiency, Qwen-3 Max achieved a score of 74.8, surpassing both Claude Opus 4 and DeepSeek V3.1. Moreover, Alibaba has teased the upcoming Qwen-3 Max Thinking model, which is currently in training and has already shown “remarkable potential,” with expectations of a perfect score on reasoning benchmarks like AIME 25 and HMMT. For those eager to experience the capabilities of Qwen-3 Max, it is available for free through the Qwen app or website. Users on iOS or Android can easily access the model via the Qwen chat app by selecting it in the settings. Dive into conversations with this cutting-edge model and see how it stacks up against your preferred chatbots.