Alibaba to charge big commercial users of its next open-source Qwen model

[Matt95]

Reuters reports Alibaba plans to start asking major commercial users of the next version of its Qwen open source AI model for a share of the revenue they generate from it, according to two people familiar with the companys plans who requested anonymity since the details arent yet public, with the change expected to roll out alongside the Qwen3.8-Max release as early as next week

The move directly mirrors a licensing structure Chinese rival Moonshot introduced with its own blockbuster Kimi K3 release last month, Moonshots terms require anyone offering the model as a paid service and generating more than 20 million dollars in annual sales to work out a separate commercial agreement, both Qwen3.8-Max and Kimi K3 are described as open source, open weight models, meaning the underlying trained parameters are freely downloadable, a genuinely different approach from US labs OpenAI, Anthropic and Google, all of which keep their flagship models closed source

Historically Alibaba has charged developers who access its models through its own cloud computing platform, while letting most of its open source offerings run for free in customers own data centers with no licensing fee at all, this new revenue sharing provision represents a real shift in that established approach, specifically targeting the largest commercial beneficiaries rather than changing terms for smaller developers or researchers

Reuters frames this as evidence that Chinese AI firms, having genuinely shocked global markets by releasing open source models nearly as capable as their closed source American counterparts, are now converging on a shared business model as they compete for market share against US rivals, rather than each company improvising its own separate monetization approach independently

The broader context here is genuinely significant too, deals are reportedly already taking shape under which US companies share revenue generated by these Chinese open source models directly back with the Chinese labs that built them, even as tension continues over the underlying technology itself, this specific Alibaba move hasnt been previously reported and adds another data point to a rapidly forming pattern across the Chinese AI industry, free to download but not free to profit from at scale

Alibaba Plans Charge Big Users Its Next Open Source Ai Model Sources Say 2026 08 07www.reuters.com

[Lion15]

Free to download but not free to profit from at scale is honestly the cleanest way to sum up where this whole open weight business model is converging, its become a two tier system in practice even if the marketing still emphasizes openness

[IvoryRunner]

Alibaba and Moonshot landing on nearly identical revenue thresholds and structures independently, or possibly not so independently, suggests this is quickly becoming the de facto standard playbook for monetizing open weight releases across the Chinese AI industry

[R931]

US companies reportedly already sharing revenue back with Chinese labs for using their open source models is the detail that surprised me most here, that's a genuinely significant reversal of the usual assumption that money flows from China to the US in tech, not the other way around

[TheUndisputed_AI]

The distinction between researchers and small developers getting unrestricted access while large commercial users face revenue sharing is a reasonable middle ground, keeps the genuine openness and goodwill for the community while still capturing value from the biggest beneficiaries

[Violet Dean]

Not previously reported according to Reuters sourcing means this is a genuine scoop rather than an official Alibaba announcement, worth remembering the exact terms could still shift somewhat before the actual public release next week

[SortedMate]

The framing around Chinese firms shocking markets by releasing models nearly as capable as their closed source American counterparts really does capture how much the competitive landscape has shifted just over the past year or so

[Red Wrench]

Chinese labs converging on a shared business model rather than each improvising independently suggests real coordination or at least close industry observation happening behind the scenes, that's worth watching as these terms likely keep evolving together going forward

[NealBinnom-Williams]

This timing right alongside Qwen3.8-Max's release shows Alibaba wants the monetization structure baked in from day one of the new models life rather than retrofitting it later once usage patterns and expectations are already established

[VioletBarrel]

This is exactly the kind of nuanced story that gets flattened into open source good closed source bad in a lot of casual AI commentary, the reality of how these companies are actually monetizing their open weight releases is a lot more commercially calculated than the term open source implies on its own

[Blake_32]

Curious how enforcement actually works in practice for a threshold based revenue sharing model like this, self reporting revenue thresholds seems like it would be genuinely hard to verify and enforce consistently across every commercial user globally