inclusionAI: Ant Group's open AI lab
inclusionAI is the open-source home of the artificial general intelligence effort at Ant Group, the Chinese company behind Alipay. It releases three complementary model families as open weights: Ling for direct answers, Ring for reasoning and Ming for multimodal tasks.
Data as of October 22, 2026
Key figures
| Country | China |
|---|---|
| Headquarters | Hangzhou |
| Type | Research lab |
| Parent company | Ant Group |
| Weights policy | Open weights |
| Model families | Ling, Ring, Ming |
| Official site | huggingface.co |
| Models tracked here | 4 |
The open lab of a fintech giant
Ant Group, Alibaba’s sister company best known for the Alipay payment app, is headquartered in Hangzhou, China. Its large language model program, called Bailing, publishes its work under the inclusionAI banner. Its motto, “AI built by everyone, for everyone”, sums up the approach: making models available to the community rather than keeping them for in-house products.
Three complementary families
inclusionAI’s lineup is organized into three series:
- Ling: instant language models with no thinking phase, designed to answer quickly and cheaply;
- Ring: reasoning models that think before answering and target long tasks, especially coding and AI agent work;
- Ming: multimodal models built on Ling, able to handle images, audio and speech (recognition and synthesis).
The flagship Ling and Ring models are among the largest open models ever released. They use a mixture-of-experts design: only a fraction of the model works on each generated word, which keeps compute costs down. Smaller versions exist for more modest needs.
inclusionAI also publishes research tools, for example for reinforcement learning or for agents that operate a graphical interface (UI-Venus).
A fully open policy
This is inclusionAI’s defining trait: the organization states that all of its projects are released under the MIT or Apache 2.0 license. Its models are therefore open-weight models in the full sense: downloadable, modifiable and usable commercially. They are available on Hugging Face, ModelScope and GitHub.
In practice, the largest models require clusters of servers with high-end graphics cards. For a quick test, a platform that hosts them is the better choice.
How to get access
You have two options: download the weights and serve them with SGLang or vLLM, or use the third-party platforms the organization mentions, ZenMux and OpenRouter. To see where Ling and Ring stand among other open models, visit our open-source models page. The profile of Alibaba, a close relative of Ant Group, covers another major open family: Qwen.
inclusionAI (Ant Group) models tracked on this site
| Models | Released | Capability index | Price (input / output) |
|---|---|---|---|
| Ling 3.1 Flash | September 29, 2026 | — | $0.075 / $0.23 |
| Ling 3.0 Flash VL | September 10, 2026 | — | $0.06 / $0.18 |
| Ling 3.0 Flash Fin | August 27, 2026 | — | $0.075 / $0.22 |
| Ling 3.0 Flash | July 23, 2026 | — | $0.06 / $0.18 |
Frequently asked questions
How is inclusionAI related to Ant Group?
inclusionAI describes itself as the home of Ant Group's AGI (artificial general intelligence) initiative. It is the name under which the group publishes its models and research tools on Hugging Face, ModelScope and GitHub.
What is the difference between Ling and Ring?
Ling covers the instant models, which answer directly without a visible thinking phase. Ring covers the thinking models, trained to reason at length before answering. Both share the same technical foundation.
Are inclusionAI models free to use?
The organization states that all its projects are released under the MIT or Apache 2.0 license, two permissive licenses that allow commercial use. It is still worth checking the license shown on each model card.
How can I try Ling or Ring without a server?
inclusionAI points to third-party platforms such as ZenMux and OpenRouter for trying its hosted models. Otherwise, you need to download the weights and serve them with SGLang or vLLM, which takes substantial hardware for the largest models.
Data sources
- huggingface.co
- github.com
- github.com
- huggingface.co
- Models.dev — Model data: models.dev (open-source project)
- Epoch AI — AI Benchmarking Hub — Epoch AI, “AI Benchmarking Hub”, epoch.ai, CC BY 4.0
Data as of October 22, 2026
Report an error
Does a figure, a date or a fact look wrong? Tell us: every report is checked.