Beijing-based AI start-up ModelBest has closed a C+ Series funding round, bringing its total fundraising this year to 5 billion yuan, or about $740 million, according to investor Shanghai Mejoy Capital. The round valued the company at 20 billion yuan, although the size of the latest financing was not disclosed. The deal underscores growing investor interest in technology that can run powerful artificial intelligence tasks directly on devices such as smartphones and laptops, rather than relying entirely on remote data centers.
ModelBest is one of the more prominent players in on-device AI, a field often described as edge AI or local AI. Its funding comes as AI developers rethink where computation should happen. Many current AI features already run locally on phones, but more demanding tasks are typically routed through cloud facilities. That approach can add latency, increase costs and raise privacy questions because user data must travel to remote servers.
A shift away from cloud-only AI
The industry is not abandoning large cloud models. Companies continue to develop models with trillions of parameters, and those systems remain central to advanced AI services. At the same time, developers are building smaller, more efficient models and techniques that allow increasingly powerful tasks to run on the device itself. The goal is not always to replace cloud AI but to divide work intelligently between local hardware and remote infrastructure.
On-device processing can be especially attractive for tasks that need immediate responses, such as voice assistants, translation, photo editing, text prediction and personal productivity tools. When a model runs locally, it can often respond faster and continue working without a network connection. Local processing can also keep sensitive data on the handset or laptop, reducing the amount of information sent to third parties.
The trade-off is that smartphones and laptops have limited computing power, memory and battery life compared with data centers. Running large models locally requires compact architectures, efficient training methods and hardware acceleration. This is why the latest wave of edge AI is focused on small but capable models, as well as specialized chips that can handle AI workloads without draining too much power.
ModelBest and Huawei chips
ModelBest has been positioning itself at the center of this transition. In May, the company said its billion-parameter on-device MiniCPM5-1B model was trained entirely on Huawei’s Ascend chips. The model achieved top benchmark scores on Artificial Analysis’ index for open-weights models under 2 billion parameters. That detail matters because it shows how China’s AI ecosystem is trying to build a more self-reliant stack, from chips to models to applications.
In China, edge AI is seen as a way to take advantage of locally produced AI chips. These chips are often better suited to training and running lighter-weight models than the most advanced frontier products. For Chinese developers, on-device AI offers a practical route to deploy AI features on consumer hardware while working within hardware and supply-chain constraints. It also creates opportunities for device makers, chip designers and software companies to collaborate on optimized systems.
The funding round suggests investors believe on-device AI can become a large market. A valuation of 20 billion yuan places ModelBest among the more closely watched AI start-ups in the country. The participation of Shanghai Mejoy Capital, which announced the round on social media, also signals that financial backers are looking beyond cloud-based model training and toward deployment at the edge.
StepFun launches edge models and a phone
Shanghai-based StepFun, which is backed by Tencent Holdings, has taken a similar bet. Last week, the company announced Step Edge, a family of on-device AI models designed to offer low latency, local data processing and the ability to handle simple tasks without a network connection. Those features align with the core promise of edge AI: faster responses, greater privacy and more reliable performance in places where connectivity is limited.
A day later, StepFun announced the StepX Neo, a smartphone designed around autonomous AI agents. According to the company, the agents are built to interpret complex user requests and carry out tasks independently. StepFun did not provide a launch date or pricing for the device. The announcement nonetheless illustrates how AI companies are moving beyond software and into hardware, hoping to design devices that make local AI a central part of the user experience.
The idea of an AI agent phone is still emerging. Instead of opening separate apps for each task, a user might ask the device to plan a trip, organize messages, edit a document or coordinate multiple services. Such features would require a combination of local models, cloud services and secure integration with personal data. Running part of that workload on the device could make interactions faster and more private, while more complex requests might still be handled in the cloud.
Apple’s on-device strategy
Apple is pursuing a parallel strategy. The company is relying on third-party providers such as Google and OpenAI for high-end tasks while also working on on-device models. Last month, Apple announced the 3 billion parameter AFM 3 Core and a more advanced 20 billion parameter model that activates between 1 and 4 billion parameters based on user tasks. Both are designed to run on devices.
That approach reflects a hybrid model for AI. A device can use a smaller local model for routine requests, then call on a larger cloud model when a task demands more reasoning or knowledge. The local model can handle quick actions, preserve privacy and reduce network dependence, while the cloud model can deliver more advanced capabilities. The challenge is making that handoff seamless and predictable for users.
Apple is also evaluating technology from Silicon Valley firm PrismML, which claimed last week to have successfully run Alibaba’s 27 billion-parameter Qwen 3.6 model locally on an iPhone. If such demonstrations can be translated into reliable consumer features, they would mark a significant step for edge AI. Running a model of that size on a phone would require aggressive optimization, memory management and hardware acceleration, and it remains to be seen how well it performs in everyday conditions.
Alibaba and Baidu are providing technology for Apple’s AI products in China, the companies said last week. The partnerships highlight how global device makers must adapt their AI strategies to local markets, where regulatory requirements, language needs and available services can differ sharply. For Apple, China is both a major market and a complex operating environment, making local partners important for AI features.
Why investors are watching edge AI
The appeal of edge AI extends across the technology industry. Cloud AI can be expensive at scale because every request consumes data center compute, bandwidth and energy. Moving some tasks to devices can reduce those costs and ease pressure on infrastructure. It can also improve reliability when networks are congested or unavailable. For device makers, on-device AI can become a selling point, differentiating phones and laptops that offer faster, more private features.
For start-ups, the edge creates new opportunities to compete. Instead of trying to match the largest cloud models parameter for parameter, smaller companies can focus on efficient models, specialized hardware support and tight integration with devices. ModelBest’s work with Huawei chips and StepFun’s launch of edge models and a phone are examples of that strategy. They are not only building models but also trying to shape the systems in which those models run.
The market still faces obstacles. Developers must manage memory constraints, battery use and thermal limits. They must ensure that local models remain accurate and safe while operating with less compute. Device makers must decide which tasks should stay on the phone and which should be sent to the cloud. Regulators and consumers will also ask how personal data is handled when AI processing moves closer to the user.
Even so, the direction of travel is clear. AI companies are no longer treating the cloud as the only place for serious computation. They are designing smaller models, optimizing chips and building devices that can handle meaningful AI workloads locally. The latest funding for ModelBest and the new products from StepFun show that investors and developers see edge AI as a practical next stage, not just a research experiment.
As smartphones and laptops become more capable, the boundary between local and remote AI will keep shifting. The most successful products may be those that combine both approaches, using on-device models for speed and privacy while drawing on cloud systems for advanced tasks. For now, the race is on to make powerful AI feel immediate, personal and available wherever a device is used.
Source: Silicon UK News