Alibaba unveiled its largest artificial intelligence model to date on Monday. The announcement sent the company's shares up 7% in Hong Kong trading.
At the same time, a research firm said DeepSeek's newest model offers pricing that is more than 100 times cheaper than Anthropic's Claude Fable 5. The two moves show how fast Chinese tech firms are pushing forward in AI development.
Both companies are racing to build stronger systems without making them too costly to run. That balance has become the main battleground for Chinese AI firms this year.
Alibaba's New Model
The new model is called Qwen3.8-Max. It has 2.4 trillion parameters, which are the numerical settings a model learns from data to recognize patterns and generate answers.
That puts it close to rival Moonshot AI's Kimi K3, which launched last month with 2.8 trillion parameters. A higher parameter count does not always mean a better model, but it is widely used as a way to measure the scale of an AI system.
Qwen3.8-Max quickly climbed AI leaderboards after it was revealed on the platform Arena.AI. It became the top-ranked Chinese model for text tasks, though it still trails Claude Fable 5 and three Opus models, all made by Anthropic.
For tasks involving images and other visual material, Qwen3.8-Max ranked second worldwide. Only a Claude Fable 5 version scored higher.
The model can process up to 1 million tokens at once. Tokens are small chunks of text, and a higher number means the model can handle longer documents, bigger codebases, or lengthy files in a single pass.
Alibaba said the model finished a software engineering project in 16 days. The company plans to fully release it next week.
Qwen3.8-Max uses a design called mixture-of-experts. This splits tasks among smaller specialized parts of the system instead of running the whole model for every request.
Only 95 billion of its parameters are active at any one time. This design cuts costs and speeds up response times.
DeepSeek's Low-Cost Model
DeepSeek also made news this week with its V4-Flash model. A research firm said it is the cheapest model to run among major AI systems available today.
Its price is more than 100 times lower than Anthropic's Claude Fable 5. That gap highlights how some Chinese firms are focusing on affordability rather than raw power.
Analyst Lian Jye Su of research firm Omdia said many business tasks do not need the most advanced model on the market. Companies often just need something reliable and low cost.
Su said this is why open-weight models have found a strong market among developers. Open-weight models let developers download and adjust the underlying system themselves.
This is different from companies like OpenAI, Anthropic and Google, which keep their models closed. Chinese firms have leaned into open access as a way to compete globally.
Both Qwen3.8-Max and V4-Flash reflect this same strategy. They aim to offer strong performance at a fraction of the cost of Western competitors.
Alibaba's model is set for full release next week, giving developers a chance to test its claims directly.
