Alibaba has unveiled Qwen3.8-Max, describing it as the company’s largest and most capable Qwen AI model to date, arriving close on the heels of a similarly massive model from domestic rival Moonshot AI. The announcement pushes Alibaba’s AI model lineup further into direct competition with both Chinese and Western developers.

Background

Alibaba’s Qwen AI family has expanded rapidly over the past year as Chinese technology companies race to build increasingly powerful large language models without letting operating costs spiral out of control. The lineage stretches back to Qwen3-Next, an efficient architecture Alibaba introduced with a mixture-of-experts design built for long-context understanding on relatively modest hardware.

That efficiency-first approach carried into February, when Alibaba launched Qwen 3.5 models built for what the company called the “agentic AI era.” At the time, Alibaba said Qwen3.5 was 60 percent cheaper to run and eight times better at handling large workloads than its predecessor, while adding the ability to independently carry out tasks across mobile and desktop apps.

Since then, Alibaba has kept up a rapid release schedule, following Qwen3.5 with Qwen3.7-Max in May and now Qwen3.8-Max this week. Each new Alibaba AI model upgrade has aimed to close the gap with both domestic rivals like Moonshot AI and DeepSeek, and international leaders in the field.

Details

Qwen3.8-Max carries 2.4 trillion parameters, the numerical settings a model learns from data and uses to recognize patterns, generate answers, and carry out tasks. That places it just behind Moonshot’s Kimi K3 model, which launched last month with 2.8 trillion parameters, though Alibaba has said Qwen3.8-Max ranks higher than Kimi K3 on several independent benchmarks.

The new model uses a mixture-of-experts design, the same efficiency principle behind Qwen3-Next, meaning only a fraction of its total parameters activate for any single task. This keeps computing costs down while still allowing the system to handle text, images, and video, along with context windows reaching up to 1 million tokens at once.

Alibaba has positioned Qwen3.8-Max as “second only” to Anthropic’s Claude Fable 5 based on its own internal evaluations, though outside coverage notes this framing hasn’t yet been backed by a full independently published benchmark table. On Arena.AI’s leaderboard for models analyzing images and visual material, Qwen3.8-Max ranked second globally, again trailing only a Claude Fable 5 variant.

Reports describing Qwen3.8-Max as a new Chinese AI video generator should be read with some caution. While early coverage suggests the model can process video input as part of its multimodal capabilities, Alibaba has not published a complete spec sheet confirming full video generation features, and some reporting treats the extent of its video and image-generation abilities as still unconfirmed.

For developers, Alibaba has said it plans to release the model’s weights for public download, meaning Qwen 3.5 Huggingface-style open access is expected to extend to the new model as well, following the pattern Alibaba set with earlier Qwen releases. The company said Qwen3.8-Max is due to become generally available next week through Alibaba Cloud’s Model Studio platform.

Quotes

Alibaba described Qwen3.8-Max in its announcement as the company’s largest and most capable AI model to date, framing the release as a direct response to intensifying competition among Chinese AI labs following Moonshot AI’s Kimi K3 launch.

According to Bloomberg’s reporting on the release, Qwen3.8-Max outperformed Kimi K3 on a number of benchmark tests, though the outlet noted that a higher parameter count does not automatically translate into better real-world performance.

Industry analysts covering the Chinese AI market point to the rapid pace of releases, from Qwen3-Next through Qwen3.5 and now Qwen3.8-Max, as evidence of just how aggressively Chinese firms are competing for technical leadership, even as questions remain about independently verified benchmark claims.

Impact

The release carries real financial weight. Alibaba’s shares climbed more than 6 percent in Hong Kong trading following the Qwen3.8-Max announcement, reflecting investor enthusiasm for the company’s continued AI investment following its previously announced plan to spend up to 380 billion yuan, or roughly $56 billion, on AI infrastructure over three years.

Globally, the release intensifies pressure on other major AI developers, both in China and internationally. Alibaba’s decision to release model weights publicly continues to position Qwen AI as one of the most significant forces in open-weight AI development, a space Chinese firms have increasingly come to dominate compared to more closed approaches favored by some Western labs.

The back-and-forth between Alibaba and Moonshot AI also reflects a broader trend reshaping China’s AI sector, where companies are racing not just to build the largest models but to do so while keeping costs manageable enough for widespread developer adoption, a balance Qwen3-Next’s efficient architecture was specifically designed to strike.

Conclusion

With Qwen3.8-Max expected to reach general availability next week, attention will turn to how the model performs once independent developers and researchers can test it directly, rather than relying on Alibaba’s own benchmark claims. Given the pace of releases so far this year, more updates to the Qwen AI lineup, and continued competition from Moonshot AI and other Chinese labs, seem likely before the year is out.

Frequently Asked Questions

What AI model does Alibaba make?

Alibaba develops the Qwen family of AI models through its Qwen Team, which has grown into one of the most extensive lineups in the industry, spanning efficient open-weight releases like Qwen3-Next to flagship closed-preview systems like Qwen3.8-Max. The company offers models suited to different needs, including versions optimized for coding and agentic tasks, multimodal models that handle text, images, and video, and lighter models designed to run on more modest hardware. Qwen models are available both through Alibaba’s own Qwen Chat consumer app and through Alibaba Cloud’s Model Studio platform for enterprise developers.

Which is the newest AI model?

As of this week, Alibaba’s newest model is Qwen3.8-Max, a 2.4 trillion-parameter system unveiled as the company’s largest and most capable release to date. It follows closely behind Moonshot AI’s Kimi K3, a 2.8 trillion-parameter model that launched just last month, making the two among the most recent large-scale AI releases from Chinese developers. Given how quickly this space moves, with new flagship models arriving every few months, it’s worth checking current coverage for the very latest release, since “newest” in this fast-moving industry can shift within weeks.

What is the newest AI model released by China?

The most recent major AI model release from China is Alibaba’s Qwen3.8-Max, unveiled this week as the company’s largest system yet, arriving shortly after domestic rival Moonshot AI released its own large-scale Kimi K3 model. Chinese AI developers, including Alibaba, Moonshot AI, and DeepSeek, have kept up an unusually fast release schedule throughout the past year, each racing to demonstrate technical leadership through massive parameter counts, competitive benchmark scores, and open-weight availability that appeals to the global developer community. This rapid cadence reflects Beijing’s broader ambitions to close the gap with, and eventually surpass, leading AI developers in the West.