Alibaba Launches Biggest AI Model Yet Amid DeepSeek’s Low-Cost AI Push
China’s AI battle heats up as Alibaba launches its biggest and most advanced AI model with 2.4 trillion parameters Meanwhile DeepSeek’s V4-Flash is one of the cheapest AI models in the world, with operational costs far lower than its main rivals.
On August 3, China's Alibaba introduced its most advanced artificial intelligence model to date. Contrarily, a research agency found that DeepSeek's newest product offers more than 100 times lower prices than Claude Fable 5 from Anthropic. Both breakthroughs demonstrate how fast Chinese IT companies are progressing in the field of artificial intelligence.
These companies are in a hotly contested race to develop increasingly powerful systems without driving up their operating costs to unsustainable levels. As they compete for developers throughout the world, two Chinese models—the Qwen3.8-Max from Alibaba and the V4-Flash from DeepSeek—highlight the country's dedication to open-weight models.
Features of Alibaba’s Qwen3.8-Max
The new Qwen3.8-Max from Alibaba soared to the top of AI model capability rankings soon after its unveiling. Parameters are the numerical settings that a model learns from data in order to identify patterns, create answers, and perform tasks; this model includes 2.4 trillion parameters. Launched last month, Moonshot AI's Kimi K3 has 2.8 trillion parameters, putting it not too far behind its domestic rival. A better model is not necessarily one with a larger parameter figure.
But now everyone is keeping an eye on it since it shows how big the data and computational power underlying AI systems are. Qwen3.8-Max debuted in Arena.AI, a crowdsourced platform for comparing models. It quickly surpassed all other Chinese text models, including Anthropic's three Opus versions and Claude Fable 5, but it is still behind the top three. On the Arena.AI's global leaderboard for visual material analysis models, Qwen3.8-Max was second only to a Claude Fable 5 variation.
DeepSeeK Most Affordable Models
According to Artificial Analysis, a San Francisco-based company, DeepSeeK's V4-Flash costs $0.14 for every million input tokens and $0.28 for every million output tokens. V4-Flash was estimated by Artificial Analysis to cost 3 cents each test, while Kimi K3, OpenAI's GPT-5.6 Sol, and Claude Fable 5 all cost 86 cents, $1.86, and $3.15, model needs to process and produce in order to finish a task, the respectively. By taking into consideration the quantity of data a comparison offers a more accurate valuation metric than price alone.
Even if the model's headline price is modest, it could end up costing a lot of money if it takes a long time to get the answer. Among Chinese AI projects, DeepSeek used to get the lion's share of attention. Nevertheless, the company was soon overwhelmed by a slew of domestic competitors, including not only other startups like Z.AI, Moonshot, and MiniMax, but also technological giants like ByteDance and Alibaba. All of them are competing with US tech companies to attract corporations looking for affordable solutions to implement AI on a large scale.
