InclusionAI

InclusionAI is an open-source artificial general intelligence research initiative established by Ant Group and operates as an internal Ant Group research division based in Hangzhou, China. It develops and openly releases large language models, multimodal models, reinforcement learning systems, agent frameworks, and evaluation platforms. The initiative was founded in 2025, and its work is organized around open-weight releases, with model weights, training recipes, and code distributed primarily through Hugging Face, ModelScope, and GitHub.

Its model families include Ling for general-purpose text, Ring for explicit reasoning and agentic workflows, Ming for multimodal any-to-any generation, and LLaDA for diffusion-based language modeling. Ling-2.6-1T is a trillion-parameter model with roughly 63 billion activated parameters, a one-million-token context window, and a hybrid architecture combining Multi-Latent Attention with linear attention. Ring-2.6-1T is a trillion-parameter mixture-of-experts reasoning model with about 63 billion active parameters per token, tuned for coding agents, tool use, and long-horizon multi-step execution.

Ming-Flash-Omni-2.0 is described as the first open model to unify speech, audio, and music generation within one architecture. LLaDA 2.1 adds token-editing acceleration. InclusionAI also develops AReaL, an asynchronous reinforcement learning framework jointly developed with Tsinghua University's Institute for Interdisciplinary Information Sciences, and operates the Inclusion Arena live evaluation leaderboard.

Through open releases under permissive licenses, most often MIT, InclusionAI lets researchers and developers use, reproduce, and build on its models rather than access them only through a hosted API. It publishes technical reports, training infrastructure code, reproducibility notes, and benchmark data, and contributes to community projects such as vLLM and SGLang for efficient inference deployment. Its stated principles are open by default, ecosystem oriented, and start small.

The initiative is led at the executive level by Zhengyu He, Ant Group's chief technology officer, while Richard Bian leads product and growth. Its team includes engineers and researchers who previously worked at OpenAI, Google, and Meta's FAIR lab, alongside Ant Group's machine learning organization. InclusionAI positions itself as a community-driven hub for projects from Ant Group's research and engineering teams, with the stated mission of treating AGI as a shared global milestone rather than a proprietary asset.