Alibaba has released its largest AI model yet, Qwen-3-Max, positioning itself alongside top competitors like OpenAI and Google DeepMind in the race for advanced artificial intelligence. The new model, Qwen-3-Max-Preview, surpasses 1 trillion parameters, marking Alibaba’s first AI system to reach this scale. It was launched on Alibaba’s cloud services platform and the AI marketplace OpenRouter on Friday. This expansion is part of the company’s ongoing Qwen3 series, which started in May with models ranging from 600 million to 235 billion parameters.
Parameters in AI models represent the system’s learning capacity. A higher number of parameters generally indicates stronger intelligence but requires more computing resources for training and operation. By comparison, OpenAI’s GPT-4.5 is estimated to have between 5 and 7 trillion parameters, making it one of the largest models globally. Qwen-3-Max-Preview is a text-only model that Alibaba says outperforms its prior Qwen3-235B-A22B-2507, released in July. Internal tests show improvements in areas such as Chinese-English text understanding, complex instruction processing, handling open-ended tasks, multilingual capabilities, and tool invocation. The model also reportedly surpassed MoonShot AI’s Kimi K2, Anthropic’s Claude Opus 4 (non-reasoning), and DeepSeek V3.1 in five benchmark tests, though the results have not been published in an official technical report.
Alibaba’s Qwen series has propelled the company to the forefront of the global open-source AI community. Its previous models have been downloaded more than 20 million times, with over 100,000 derivative models created on the developer platform Hugging Face. However, Qwen-3-Max-Preview has not been open-sourced and is currently accessible only through official channels. Similarly, Qwen2.5-Max, released in January, was not open-source. An Alibaba AI engineer, Binyuan Hui, confirmed on social media that a “thinking” version of the model is in development, hinting at future enhancements.
Access to Qwen-3-Max-Preview on Alibaba Cloud follows tiered pricing. Input tokens start at US$0.861 per million, and output tokens cost US$3.441 per million, making it one of the most expensive models in the Qwen series for API access. By comparison, the non-thinking Qwen3-235B-A22B-2507 charges US$0.287 per million input tokens and US$1.147 per million output tokens. MoonShot AI’s Kimi K2 charges US$0.60 per million input tokens and US$2.50 per million output tokens.
Alibaba has pledged 380 billion yuan (US$52 billion) for AI infrastructure over the next three years, surpassing its total AI investment over the past decade. The company is already seeing significant returns, with AI-related products achieving triple-digit growth for eight consecutive quarters, according to the latest financial results. Alibaba’s ambitious AI initiatives reflect the company’s commitment to scaling its capabilities and competing with global leaders. With Qwen-3-Max, it is clear that the Hangzhou-based tech giant is serious about expanding its presence in the trillion-parameter AI arena.
