StepAI CEO Jiang Daxin
AsianFin -- StepAI, a prominent AI startup, is gearing up for the release of its full-powered inference model, Step R1, within the next two to three months, according to founder and CEO Jiang Daxin.
The anticipated launch is expected to make waves in the artificial intelligence industry, alongside the rollout of a more advanced Step image editing model.
In a recent conversation with AsianFin, Jiang emphasized the importance of breakthroughs in AI models preceding their commercialization. He noted that integrating multi-modal understanding and generation is crucial for StepAI's mission to build a "world model," advancing toward Artificial General Intelligence (AGI) and intelligent agents.
"We have no fear," Jiang said, adding that the path to AGI is clearer than ever. "We are confident in our ability to develop models that integrate diverse forms of intelligence, such as visual and spatial intelligence, which will ultimately lead to AGI."
Jiang also discussed the lessons learned from the launch of DeepSeek, highlighting that the traditional model of traffic investment is no longer valid in the AI era. He pointed out that products like ChatGPT, despite not opening up traffic, still reached significant user bases, signaling a shift in the understanding of AI product growth.
"AI product traffic growth is not reliant on traditional internet models," he explained. "This needs to be re-evaluated, as evidenced by the performance of AI products like DeepSeek, Nezha, and Black Wukong."
Looking ahead, Jiang outlined three major AI technology directions that StepAI is focused on:
1. Pre-trained foundational models with reinforcement learning to enhance reasoning capabilities in models.
2. Integrated understanding and generation in the visual domain using a single model, making the generated content meaningful and contextually relevant.
3. The evolution of AI agents, or intelligent entities, that extend into the physical world, including applications in autonomous driving and humanoid robots.
Jiang expressed particular enthusiasm for the development of intelligent terminal agents, emphasizing that these agents must understand their users' environment and context to effectively assist in completing tasks. He highlighted that devices like AI glasses and smartphones are critical for collecting environmental data, thereby improving AI models' contextual awareness.
StepAI, founded in April 2023, has rapidly emerged as a key player in the race to develop foundational AI models. The company has already released 22 self-developed foundational models across multiple modalities, earning the title of "Multimodal King" in the industry. With a heavy focus on multimodal AI, StepAI has partnerships with companies such as Geely Auto, OPPO, and Zhiyuan Robotics, and continues to explore cutting-edge applications in smart terminals.
Looking at the broader competitive landscape, Jiang noted that major international players like OpenAI and Google are in the first echelon of AI development, while StepAI's differentiator lies in its extensive multimodal capabilities. This focus on multimodal development, he believes, positions StepAI well for future success in both foundational research and generational AI applications.
StepAI's commitment to advancing AGI remains unwavering. "We will continue to develop foundational large models as they are crucial to the realization of AGI," said Jiang. He also pointed out that the application of AI models and their development go hand in hand, with applications providing valuable data to improve and push the boundaries of foundational models.
In terms of financial growth, StepAI secured significant backing in December 2024, closing a Series B funding round with several hundred million dollars. Key investors include Tencent Investment, Shanghai State-Owned Capital Investment, and Qiming Venture Partners, positioning the company to scale its ambitious vision for AGI and intelligent agents.
As StepAI progresses in its mission, Jiang remains optimistic about the future of AI, particularly in the development of intelligent entities that will blur the lines between digital and physical worlds. The company is now charting a differentiated course toward building an AI ecosystem that spans from foundational models to AI agents, and from cloud to edge computing.