StepFun

Stepfun is focused on advancing artificial general intelligence (AGI) through the development of self-developed, unified multimodal models. These models are designed to process and understand various inputs including language, images, video, and speech. Since March 2024, the company has released 11 self-developed base models, including the Step-1 and Step-2 language models. Notably, its Step-1V is a multimodal large language model with over 100 billion parameters, and the Step-2 model, built on a Mixture of Experts (MoE) architecture, is a trillion-parameter model, the first of its kind from a Chinese startup. This model has achieved a high ranking on Livebench, a global AI model benchmarking platform.

The company's business model centers on providing its advanced models to AI research and development users. The technology is also accessible through an API platform for developers and researchers and has been integrated into a consumer-facing application called "Yue Wen". Stepfun has formed strategic partnerships to apply its technology in various sectors, including finance, content creation, and smart terminals with companies like Zhongwen Online, Honor, and OPPO. In a joint venture with Jiemian Finance, a subsidiary of the Shanghai Newspaper Group, Stepfun co-founded Caiyue Xingchen to develop large model applications specifically for the financial industry. In late 2024, the company secured a Series B funding round of several hundred million dollars, with backing from investors such as Fortera Capital, Tencent Investment, Qiming Venture Partners, and 5Y Capital. This capital is intended to further research in foundational models, enhance multimodal reasoning, and expand consumer application scenarios.