Alibaba Doubles Down on Open Source at ModelScope DevCon
Alibaba confirmed its commitment to open-sourcing Qwen, Wan, and new model families at ModelScope DevCon in Nanjing. Here's what was announced.
Vlad Makarovreviewed and published
At a time when Chinese AI labs are increasingly going proprietary, Alibaba just went the other direction. At this weekend's ModelScope DevCon in Nanjing, Jingren Zhou — head of Alibaba's AI division — confirmed the company will keep open-sourcing Qwen, Wan, and several entirely new model families.
What Was Announced
The phrasing from the official ModelScope account was deliberate: "Alibaba persists in open-sourcing the Qwen, Wan, and other series of models, advancing together with the ModelScope community." The verb choice — "persists" — reads like a direct response to recent speculation that the Qwen team might pivot to closed development after several senior researchers departed earlier this year.
Among the new model families teased at the event: Qwen-T (details still scarce), Qwen-TTS for speech synthesis, and Qwen-Image for visual generation. None of these have public release dates yet, but the announcement itself is the signal that matters — Alibaba is betting that the open-source playbook still works.
This comes barely a month after Alibaba launched Qwen 3.5, its most capable agentic model to date, which quickly climbed to the top of LMArena. The Qwen ecosystem now spans text, vision, code, and omni-modal models, making it the most comprehensive open-source model family from any single organization.
Why This Matters
The open-source AI landscape is at a crossroads. MiniMax just released M2.7 as proprietary. Z.ai's GLM-5 Turbo is closed-source. Even within Alibaba, there were credible rumors of a strategic shift. Against that backdrop, Alibaba's public recommitment to open weights is significant — not just for developers who rely on Qwen models, but for the broader ecosystem that depends on Chinese labs pushing the frontier of what's freely available.
The Reddit community's reaction was enthusiastic, with the announcement pulling nearly 1,000 upvotes on r/LocalLLaMA. For developers building on Qwen 3.5 or waiting for the next generation, the message is clear: more is coming, and it will be open.
