On December 23, the rapidly rising AI company, LeapStar, which specializes in large-scale models, made headlines by successfully closing a Series B funding roundThis remarkable achievement brought in several hundred million dollars, showcasing the company's potential in the flourishing AI sectorSignificant players participated in this funding, including state-owned entities, strategic partners, and financial investorsAmong the core investors were the Shanghai State-owned Capital Investment Company and its affiliated funds, alongside prominent names such as Tencent Investment, Wuyuan Capital, and Qiming Venture Partners.

The fresh influx of capital is set to further fuel the research and development of LeapStar's foundational models, enhancing its multi-modal and complex reasoning capabilitiesThe company aims to broaden its reach into consumer-facing applications, creating a more immersive user experience through an expanded product ecosystem

This strategy mirrors a broader trend in the tech world where consumer-focused AI applications are becoming integral to daily life.

LeapStar was founded in April 2023 by DrJiang Daxin, a former global vice president at Microsoft, and operates out of Xuhui District in ShanghaiThe company boasts a remarkably talented team, including Zhang Xiangyu, one of the key authors of the renowned ResNet paper in computer vision, and Zhu Yibo, an AI systems expert with experience at industry giants like Microsoft, ByteDance, and GoogleCurrently, the majority of the company's personnel—over 80%—are dedicated to research and development, underscoring its commitment to innovation.

Within the competitive landscape of “the six small dragons” in the field of large models, LeapStar has adopted a more understated approach, refraining from publicly disclosing its financial progress in the past

• Escalating Economic Stagnation in Germany
• Rising Expectations for Federal Reserve Rate Cuts
• Signs of a Cooling U.S. Labor Market?
• The Inception and Future of Flash AI Glasses
• The Rise of Chips, AI, and Robotics

However, a significant investment initiative was launched in October with the establishment of the Shanghai Artificial Intelligence Ecological Fund, spearheaded by Shanghai State Investment Company, which boasts a staggering scale of 10 billion yuanWith the release of this fund, LeapStar appeared on Shanghai State Investment’s strategic cooperation list, alongside another Shanghai-based unicorn, MiniMax.

Interestingly, there has been speculation that a few companies among the six have opted to abandon pre-training modelsIn contrast, LeapStar remains committed to this approach, emphasizing that it has developed a comprehensive suite of models that spans from comprehension to generation, encompassing both text and multi-modal formatsThis dedication positions them as one of the few companies in China keen on advancing towards Artificial General Intelligence (AGI).

Jiang Daxin, the founder and CEO of LeapStar, has laid out a distinct technological roadmap for achieving AGI, which includes a progression from single-modality to multi-modality, culminating in unified multi-modal comprehension and generation, and eventually reaching AGI through the development of world models.

Over the past ten months, LeapStar has released 11 self-developed foundational models under its Step series, which include groundbreaking creations like the Step-1 model with 100 billion parameters, the trillion-parameter Step-2 model, various image and video generation models, and the recently introduced end-to-end audio model, Step-1o Audio

These models have consistently ranked at the top of evaluation benchmarks like LiveBench and Chatbot Arena, affirming LeapStar's innovative capabilities in the AI landscape.

With an eye on commercializing large models, LeapStar emphasizes a hybrid approach that combines self-developed technology with ecological partnerships to cater to end usersOne of their flagship products is the AI assistant “Leap Inquiry” which facilitates multi-modal interactionsMoreover, their pioneering multi-modal visual search function, “Photo Inquiry,” has become a standout feature integrated into the iPhone 16's camera control—a testament to the advancing utility and accessibility of AI in everyday applications.

Looking ahead, the firm is already witnessing dramatic growth in the API usage of its multi-modal capabilities—an increase of over 45 times in just the second half of 2024. Numerous developers are now leveraging LeapStar's technology to build consumer applications

For example, Zhao Chun, the developer behind the popular AI app “Stomach Book,” has publicly lauded LeapStar as having the highest conversion rate after conducting extensive A/B testing with various models available in ChinaAnother application that has garnered significant social media traction, the “Lyric Transformer,” also utilizesLeapStar's multi-modal large models.

To further expand its industry ecosystem, LeapStar has sought strategic partnerships across several sectors, including finance, content creation, and smart hardwareIn the realm of finance, they co-founded a large-model tech enterprise, Financial LeapStar, with the Shanghai-based media group JiemianThis partnership has produced vertical large models specifically tailored for the financial industry, like Finstep, alongside personalized wealth assistant applications termed “Little Fortune God.” In the smart hardware sector, LeapStar has established collaborations with prominent smartphone manufacturers such as Honor and OPPO.

As the second half of the year turns, there has been a noticeable slowdown in the financing progress of leading domestic large model firms