Alibaba Open-Sources Qwen 3.8 Max, a 2.4 Trillion-Parameter Model: Why the US Fears Chinese Models

On July 20, 2026, analyst Ben Thompson of Stratechery proposed that the US legally recognize AI model distillation and training-data collection as fair use. His proposal came as Alibaba released Qwen 3.8 Max, a 2.4 trillion-parameter model with open weights.

What Ben Thompson proposes

Ben Thompson proposes that the US pass a two-part law: (1) explicitly establish that collecting data to train models is fair use, and (2) bar user agreements from blocking distillation, at least for American companies. In his view, stopping distillation technically is nearly impossible, since it amounts to ordinary API requests.

Thompson points to the hypocrisy of the labs: they forbid distilling their models even though they themselves trained on unlicensed data. Such a law, he argues, would help American open models compete more effectively with Chinese ones.

• The analysis was published on July 20, 2026 (a note by Simon Willison)
• Qwen 3.8 Max: 2.4 trillion parameters, open weights
• For comparison: Kimi K3 has 2.8 trillion parameters
• In May 2026, Alibaba decided NOT to open Qwen 3.7 Max
• Xi Jinping's speech on open source: July 18, 2026

"The US should go in the opposite direction and bet on a new copyright

policy that simultaneously protects the labs and guarantees that what they have learned feeds further innovation for everyone else." — Ben Thompson, Stratechery

Why Alibaba open-sourced Qwen 3.8 Max

Alibaba released Qwen 3.8 Max with open weights, reversing its May decision not to open the previous version, 3.7 Max. Thompson ties the reversal to a July 18, 2026 speech by Chinese leader Xi Jinping, who urged the country to seize the moment to advance open source.

"We must seize this rare historic opportunity to encourage open source, openness, collaboration and sharing." —

Xi Jinping, as published by the State Council of the PRC

Alibaba's Qwen 3.8 Max has 2.4 trillion parameters, nearly as many as Kimi K3 (2.8 trillion). Developer Simon Willison ran it through his "pelican riding a bicycle" test. In its extended chain of reasoning, the model mulled "Add a helmet? No" and "Maybe a little bell? No" before finally drawing a pelican on a red bicycle.

What this means

Open weights are becoming a tool of geopolitics. While American labs fight lawsuits over training data and ban distillation, China is encouraging openness at the level of Xi Jinping's rhetoric, and Alibaba changed its strategy for its flagship model in just a couple of months. Thompson's proposal is an attempt to respond in kind: rather than banning copying, legalize it, so that what the labs have learned works for the whole industry.

Frequently asked questions

What is model distillation?

Distillation is training one model on the responses of another, obtained through its API. Thompson notes that technically it is just API requests, so blocking the process is nearly impossible.

How big is Qwen 3.8 Max?

Qwen 3.8 Max has 2.4 trillion parameters, nearly as many as Kimi K3 at 2.8 trillion. Alibaba released the model with open weights.

Why is this a reversal for Alibaba?

In May 2026, the company decided not to open the weights of Qwen 3.7 Max. A couple of months later, it opened 3.8 Max. Thompson links this to Xi Jinping's July 18, 2026 speech on open source.