Alibaba’s big AI launch looks a little desperate

ALIBABA VS DEEPSEEK

Alibaba’s big AI launch looks a little desperate

Alibaba has launched Qwen 2.5-Max, saying it beats OpenAI’s GPT-4o, Meta’s Llama 3.1-405B, and China’s fast-growing DeepSeek-V3.

The timing—on the first day of Lunar New Year—shows an urgency to respond as DeepSeek gains ground.

Since DeepSeek-V3 came out on 10 January, it has made waves with its low-cost, high-performance AI models.

The impact has been so big that even ByteDance (TikTok’s parent company) updated its own AI model just days after DeepSeek’s latest release.

In brief:

• Alibaba’s Qwen 2.5-Max joins the race to compete with OpenAI, Meta, and DeepSeek.

• DeepSeek’s cheap AI models started a price war, forcing Alibaba and others to cut prices by up to 97%.

• DeepSeek is focused on AGI, running a small team of researchers instead of a big corporate structure.

DeepSeek isn’t here to play

DeepSeek’s rise isn’t just about competition—it’s changing the AI market.

When its V2 model launched last May, it cut token processing costs to just 1 yuan ($0.14) per million tokens.

This forced Alibaba, Baidu, and Tencent to drop their prices sharply.

Despite having a small team, DeepSeek is set on developing AGI, prioritising AI progress over short-term profits.

Founder Liang Wenfeng has questioned whether big tech firms, with their high costs and strict structures, can keep up with AI’s rapid changes.

DeepSeek dropping the Uno reverse card on big tech.