Alibaba Unveils Qwen4-Max, Claims Lower Prices Than Rival Top Models
Alibaba has released Qwen4-Max, a new flagship model that the company says can keep pace with leading competing models while being offered at a significantly lower price.
Alibaba has unveiled Qwen4-Max, its new flagship language model. The Chinese technology group positions the system as a high-performance yet markedly cheaper alternative to the current top models from Anthropic, OpenAI and Moonshot AI. Specifically, the model is compared with Opus 5.5, GPT-6 and Kimi K3.1 Pro.
The model has a total of 2.4 trillion parameters, although only 71 billion parameters are active per token. This separation between total size and active computing load is a central feature of the architecture. According to the available information, Alibaba relies on a hybrid architecture that is primarily intended to reduce costs when handling long contexts. Long inputs are among the most computationally intensive tasks for large language models, which is why efficiency gains in this area are particularly relevant for both providers and users.
In the performance evaluation by the analytics company Artificial Analysis, its Intelligence Index, Qwen4-Max ranks just behind Kimi K3.1 Pro. The gap to the other comparison models is not specified in detail in the available information. Alibaba presents the model as considerably less expensive than the competitors named. However, concrete prices and the exact differences in benchmark results are not stated in the available information. An independent verification of the claimed cost advantages is also not yet available.
In parallel, Alibaba announces that it will release smaller models from the Qwen4 series as open-source versions. The timing and license terms of this release are not yet known. Whether and to what extent developers and companies can use the smaller variants therefore remains open.
Source: the-decoder.com
32
