Alibaba’s Next Qwen Model Will Charge Heavy Commercial Users a Revenue Share
Alibaba plans to require large commercial users of its next Qwen open-weight AI model to share revenue, mirroring a similar clause in Moonshot's Kimi K3 license.
Alibaba is preparing to release the next version of its Qwen open-source AI model as early as next week, and for the first time it plans to require high-revenue commercial users to hand over a portion of their earnings. According to two sources familiar with the plans, the move mirrors a clause already in the license for Moonshot's Kimi K3 model, which requires any business generating more than $20 million in annual sales from the model to negotiate a commercial agreement with Moonshot, including a revenue share of up to 30 percent.
What happened
| Detail | Fact |
|---|---|
| Expected release | Next week (as of August 7, 2026) |
| Model | Alibaba Qwen3.8-Max (open-weight) |
| Revenue trigger (Moonshot precedent) | $20 million in annual sales from the model |
| Moonshot revenue share rate | Up to 30% |
| Alibaba revenue share rate | Not yet decided; discussions ongoing |
| Kimi K3 price vs. Anthropic Fable | Approximately one-third the cost per token |
Alibaba’s Qwen3.8-Max is an open-weight model, meaning developers can download the underlying learned parameters and run or modify the system on their own hardware. That is in contrast to closed models from OpenAI, Anthropic, and Google, where the weights are never released. Until now, Alibaba has charged developers who use its models via Alibaba’s own cloud, but let most open-source users run the models in their own data centers for free.
The new license will change that for large commercial operators. Two people familiar with Alibaba’s strategy told Reuters the company plans to adopt a clause similar to the one tucked into Moonshot’s Kimi K3 license: once a business crosses $20 million in annual revenue derived from the model, it must enter a commercial agreement that includes a revenue share. Alibaba’s specific percentage has not been set.
Why this matters for businesses using open-source AI
The phrase “open source” has long implied free. That assumption is now quietly eroding for businesses that scale up. The Moonshot playbook, which Alibaba is now copying, borrows directly from the Silicon Valley freemium model: zero cost for small users, commercial terms for anyone making serious money from the technology.
Paddy Srinivasan, CEO of cloud firm DigitalOcean Holdings, which already offers Kimi K3, described it plainly: “This is a tried and tested open-source freemium model.” DigitalOcean has a commercial agreement with Moonshot, though Srinivasan declined to detail the terms. He noted that paying customers get benefits like early access to future model revisions and deployment optimization support.
Dan Fu, VP of kernels at Together AI, added that companies offering AI services earn their margin by optimizing how tokens (the basic units an AI model processes per query) are used, not just reselling raw model access. The revenue-share arrangement may cut into that margin for those crossing the $20 million threshold.
Even at the new commercial rates, Chinese models remain significantly cheaper. Kimi K3 costs about a third of Anthropic’s Fable model on listed per-token pricing. That price gap gives Chinese open-weight models a strong pull, even with a revenue share attached.
The broader shift: Chinese AI firms find a business model
Chinese AI labs have rattled the market by releasing open-weight models that approach the capability of closed U.S. models. Now they are figuring out how to monetize that attention. Chinsoft International, a Chinese IT services firm, disclosed a revenue-sharing agreement with Moonshot in a regulatory filing last month, though the percentage was not disclosed.
Meanwhile, U.S. open-source AI is also growing. Thinking Machines Lab, a startup founded by former OpenAI CTO Mira Murati, released its first open-source model last month. Lin Qiao, CEO of Silicon Valley inference firm Fireworks AI, said he sees no fundamental barrier to powerful open-source U.S. models emerging. The commercial licensing trend may well follow them too.
We covered Alibaba’s earlier Qwen 3.8-Max release and the open-weight safety concerns it raised in our earlier Qwen 3.8-Max coverage.
Our take
The “open source is free” assumption has always had fine print, and now that fine print has a dollar figure attached. For businesses using open-weight models at scale, the practical question is: are you above $20 million in annual revenue from AI-powered services? If yes, read your license carefully before your next model upgrade.
The 30% ceiling on Moonshot’s Kimi K3 is aggressive. Even if Alibaba sets a lower rate, this signals that the cost of building on Chinese open-weight models is no longer just compute. Businesses that have embedded these models deeply into products should model out what a 10 to 30 percent revenue share would do to unit economics now, before the next license drops.
If you are building AI-powered products or workflows on top of open-weight models, it is worth reviewing your exposure. Our AI integration work for clients always factors in licensing risk alongside capability, and this development is a good reason to put that conversation on the agenda sooner rather than later.
What to do about it
- Pull the current license for every open-weight model you are using commercially and look for revenue-share or commercial-use triggers.
- Estimate your annual revenue attributable to each model to check whether you are near or above the $20 million threshold.
- Wait for Alibaba’s official Qwen3.8-Max license text next week before committing to any new deployments built on it.
- Compare total cost of ownership across model providers, including potential revenue-share obligations, not just per-token prices.
Frequently asked questions
Is Alibaba's Qwen model still free to use?
For most developers it remains free. The new commercial clause only applies to businesses generating more than $20 million in annual revenue from the model, mirroring the threshold in Moonshot's Kimi K3 license.
What is the revenue share rate for Moonshot's Kimi K3?
Moonshot can require up to 30% of revenue from commercial users who exceed $20 million in annual sales derived from the Kimi K3 model, according to Reuters sources.
How does Kimi K3 pricing compare to Anthropic?
Based on listed per-token prices, Kimi K3 costs approximately one-third the price of Anthropic's Fable model for input and output tokens.
What is an open-weight AI model?
An open-weight model is one where the underlying trained parameters are publicly available for download, allowing developers to run or adapt the model on their own infrastructure without relying on a vendor's API.
