The End of Free Qwen Code
In a decisive shift that marks the end of an era for open-source artificial intelligence, Alibaba has abruptly discontinued the free tier for its highly popular Qwen Code terminal agent. Decrypt reports how the announcement, made via a terse update to the project’s GitHub repository on April 15, stated simply, “Qwen OAuth free tier has been discontinued.”
Alongside the elimination of the free tier, Alibaba drastically reduced the daily quota for free API requests from 1,000 to just 100. Users seeking to run Qwen Code in the cloud are now directed to paid solutions, such as the Alibaba Cloud Coding Plan Pro ($50/month) or third-party providers like OpenRouter and Fireworks AI.
Qwen Code is not a minor experimental project; it is a sophisticated terminal coding agent designed to rival Anthropic’s Claude Code. Powered by the Qwen3-Coder models, it supports multi-file repositories and boasts SWE-Bench scores that are highly competitive with the best paid tools globally. The decision to erect a paywall around such a capable tool underscores a fundamental change in Alibaba’s AI strategy. EastFrontier has previously reported how Alibaba is shifting its strategy toward closed-source AI models to drive revenue.
A Broader Industry Pivot to Closed Source
Alibaba’s move is not an isolated incident but part of a broader trend among Chinese AI developers, who are pivoting away from open-source strategies that fueled their initial rapid growth. Just 48 hours prior to the Qwen Code announcement, fellow Chinese AI startup MiniMax executed a similar maneuver.
MiniMax released M2.7, a massive 230 billion-parameter model that rivals Claude Opus 4.6 on coding benchmarks, initially under what it described as “MIT-style” terms, allowing unrestricted commercial use. However, within hours of the launch, the company rewrote the license to require written authorization for commercial applications. While MiniMax claimed the change was to protect against bad-faith hosting providers offering degraded versions of the model, the commercial restrictions remained firmly in place.
Similarly, Xiaomi recently shipped its MiMo v2 model under a closed-source license. According to reports from the Financial Times, Alibaba’s own Qwen team has been steadily moving toward proprietary development following the departure of key leadership figures. This shift represents a stark departure from the philosophy articulated by Alibaba Group Chairman Joe Tsai just months ago, when he championed open-source AI as a means to bring global benefits and lower costs.
The Cost of Dominance and the Pressure to Monetize
The initial embrace of open-source models by Chinese firms was spectacularly successful in driving adoption. Chinese open-source models expanded their share of global open-model usage from a mere 1.2% in late 2024 to approximately 30% by the end of 2025. Alibaba’s flagship Qwen family achieved nearly 1 billion cumulative downloads, overtaking Meta’s Llama to become the most widely deployed self-hosted model family on the planet.
However, this massive market share was built largely on free services, a model that is increasingly difficult to sustain. The immense computational costs associated with training and serving state-of-the-art AI models are forcing companies to prioritize revenue generation. As investors demand returns on massive capital expenditures, the pressure to monetize inference and cloud services has intensified.
Furthermore, the geopolitical landscape is complicating the open-source ethos. With U.S. chip export controls tightening and the strategic AI race between Beijing and Washington accelerating, giving away cutting-edge capabilities for free is facing increased scrutiny. While Alibaba’s models technically remain open source for those with the substantial hardware required to run them locally, the era of frictionless, free access to top-tier Chinese AI capabilities via the cloud appears to be drawing to a close.