Recently, ModelBest has completed a new round of financing worth hundreds of millions of yuan. This round of financing is jointly invested by Hongtai Fund, Guozhong Capital, Tsinghua Holding Jinxin, and Moutai Fund. The completion of this round of financing will further lay a solid foundation for ModelBest to build efficient large language model technology, product barriers, accelerate industry empowerment, and ecological expansion, and collaborate with upstream and downstream industries to promote the large-scale application of the “edge-side brain” in various industries.

At the beginning of 2025, the global artificial intelligence competition has entered a fierce stage, and the innovation path represented by “efficiency and low consumption” is leading the global AI transformation. ModelBest is a large language model company with “efficiency” as its first principle, focusing on creating efficient large language models that have higher performance, lower cost, lower power consumption, and faster speed under the same parameters. In January of this year, ModelBest released its first edge-side full-modal model, ModelBest MiniCPM-o 2.6. Many key capabilities such as “continuous watching, real-time listening, and natural speaking” are industry-firsts, and its overall capabilities rank among the international leading level. Previously, ModelBest MiniCPM gained global recognition for its characteristics of achieving more with less, high efficiency, and low cost. It successively achieved the edge-side equivalents of ChatGPT, GPT-4V, and GPT-4o, and was named the most downloaded and most popular Chinese large language model on Hugging Face in 2024. As of now, the cumulative downloads of the ModelBest MiniCPM series across all platforms have exceeded 10 million.

ModelBest continues to promote the expansion and popularization of edge-side large language models in various fields, integrating high-performance edge-side AI into different terminal products, thereby bringing “intelligent” changes to everyone’s work and life and winning industry recognition. It is worth mentioning that in creating intelligent cockpits, ModelBest released cpmGO, the industry’s first pure edge-side automotive assistant for intelligent cockpits with the MiniCPM model as the core, creating a “native edge-side experience” for intelligent cockpits that is fully aligned with the cloud side. Through in-depth cooperation with top automotive enterprises such as Changan Automobile, SAIC Volkswagen, and Great Wall Motor, it has achieved fixed-point installations for multiple mass-produced vehicle models. The first mass-produced vehicle with an edge-side model, the Changan Mazda MAZDA EZ-60, was officially released in April this year, marking a new stage in the commercial implementation of edge-side large language models in the automotive cockpit field. In addition, ModelBest has deeply cooperated with benchmark enterprises such as Qualcomm, Intel, MediaTek, Wutong Carlink, Desay SV, Thundersoft, and Aptiv to jointly promote the widespread implementation of edge-side native intelligent cockpits.

All along, ModelBest has adhered to the concept of “taking the path one step ahead of others and winning battles with fewer resources”, continuously leading the development of efficient large language models. As one of the few international teams that have explored their own “methodology for efficient large language models”, ModelBest discovered and proposed the Densing Law of large language models. Aiming to pursue higher knowledge density (Knowledge Density = Model Capability / Model Parameters Involved in Calculation), and with outstanding human efficiency and R&D input-output ratio, it continuously iterates the real utility of the model, leading terminal intelligence to a higher level.

Hongtai Fund stated, as investors who have long been concerned about the development of artificial intelligence, we firmly believe in the industrial transformation brought about by technological equalization and believe that “efficiency” is the first principle of large language models. The essence of intelligence is not the difference between large and small, or between edge and cloud, but the difference between high and low knowledge density. The edge side is one of the scenarios where high knowledge density can be best reflected and applied. As edge-side large language models and AI Agents gradually penetrate into the control rights of devices, human-computer interaction shifts from GUI to VUI+GUI, and the user traffic and the industrial pattern of intelligent terminals will face redistribution. In the future, we look forward to ModelBest placing intelligence closest to users, achieving autonomous learning, and leading the development of terminal intelligence.

Guozhong Capital stated, we are delighted to see the outstanding achievements and remarkable contributions made by ModelBest in promoting edge-side intelligence. As a Chinese AI enterprise focusing on promoting the development of edge-side large language model technology, it not only represents China’s innovation ability and competitiveness on the global AI stage but also reflects the unique charm of Chinese-style innovation. We are honored to participate in the development journey of ModelBest, helping it accelerate the process of AI industrialization and promoting the long-term development of edge-side intelligence.

Tsinghua Holding Jinxin Capital stated, ModelBest has always been at the forefront of innovation in edge-side large language models. With its lightweight and high-performance technology route, it promotes the widespread application of large language models on terminal devices. This “edge-side first” concept not only reflects the company’s forward-looking grasp of industry development but also demonstrates its efficient innovation ability at the technical level. At the same time, we are pleased to see that it has made in-depth deployments in more and more terminal fields such as automobiles, mobile phones, and computers. With the rapid development of the terminal intelligence industry in the future, ModelBest will continue to lead the development trend of edge-side large language models and set a benchmark for the entire industry.

Regarding the new round of financing, Li Dahai, Co-founder and CEO of ModelBest, said, now, large language models have entered an accelerated moment of industry empowerment, which requires us to have more forward-looking judgments on technology and the market and provide sufficient supply for the industry empowerment of large language models. All along, ModelBest has aimed to create efficient large language models and has repeatedly achieved leading performance at a higher level in key model indicators. After this round of financing, we will be able to further strengthen ModelBest’s barriers and continue to promote the commercialization process of large language models, contributing to the arrival of the AGI era.

ModelBest

ModelBest is an enterprise that innovates in artificial intelligence large language model technology and implements its applications. With the vision of “Intelligence Surrounding All Things”, it is committed to creating safe and inclusive general artificial intelligence, enabling AI technology to benefit thousands of households.