Alibaba
The chinese tech giant Alibaba has released a new AI model Qwen-3-Max-Preview with over a trillion parameters.
The parameters define accuracy of the AI model. However, a large number of parameters requires more computing power to adjust them and more energy.
Currently, OpenAI’s GPT-4.5 model is considered to be one of the largest in the world. It probably includes 5 to 7 trillion parameters.
Qwen-3-Max-Preview is a text-based AI model, that has surpassed the best model from Alibaba, Qwen3-235B-A22B-2507, released in July of this year.
The Chinese company has released benchmark results, according to which, Qwen-3-Max-Preview has outperformed MoonShot AI’s Kimi K2, Anthropic’s version of Claude Opus 4 without logical thinking, and DeepSeek V3.1 in five tests. However, these results were not confirmed in the official technical report.
“Qwen3-Max-Preview shows significant gains in overall capabilities, including significant improvements in chinese and english text comprehension, understanding complex instructions, completing subjective open-ended tasks, multilingualism, and tool calling. The scaling is working, and the official release will surprise you even more”, — Alibaba said.
The Qwen series of AI models has allowed Alibaba to rank first in the world in terms of popularity among open source AI systems. Alibaba’s AI models have been downloaded more than 20 million times and created 100 thousand derivative versions on the Hugging Face open source platform for developers.
Qwen-3-Max-Preview is not yet open source, but the model itself is available on official channels. The previous model Qwen2.5-Max was also not open source. According to an AI engineer at Alibaba Binyuan Hui, a version of Qwen-3-Max-Preview with logical thinking is due to appear soon.
Qwen-3-Max-Preview has a tiered pricing policy with a rate of $0.861 per million incoming tokens and $3.441 per million outgoing tokens. This makes this model one of the most expensive in the Qwen series. Compared to it, the Qwen3-235B-A22B-2507 version without logical thinking charges $0.287 per million incoming tokens and $1.147 per million outgoing tokens. The new Kimi K2 model charges $0.6 per million incoming tokens and $2.5 — per million outgoing tokens.
Alibaba plans to allocate $52 billion for AI infrastructure over the next three years. The media has also reported, that the chinese tech giant is developing its own AI processor in an effort to reduce dependence on Nvidia chips.
Source: Interesting Engineering
Контент сайту призначений для осіб віком від 21 року. Переглядаючи матеріали, ви підтверджуєте свою відповідність віковим обмеженням.
Cуб'єкт у сфері онлайн-медіа; ідентифікатор медіа - R40-06029.