TMTPOST — On-device artificial intelligence developer ModelBest [Mianbi Zhineng] closed its latest funding round on Wednesday, pushing its cumulative first-half 2026 financing past 5 billion yuan ($691 million). The specialized funding round—which elevates the startup’s post-money valuation beyond 20 billion yuan ($2.76 billion)—secured capital from an array of state-backed investment funds, state-owned enterprises, automotive manufacturers, and private financial institutions. The Tsinghua University NLP Lab spin-off will deploy the proceeds to scale its proprietary "MiniCPM" edge-computing model matrix, accelerate hardware-level neural engine optimizations, and expand its developer ecosystem. By integrating its lightweight models natively into smartphones, automotive cockpits, and personal computers, the Beijing-based enterprise aims to commercialize low-latency local inference processing without relying on expensive cloud data centers. Capital allocation within the artificial intelligence sector is shifting toward cost-efficient edge execution as enterprise clients push to bypass astronomical cloud-based computing overhead. Since the operational costs of running multi-billion parameter models via external servers remain prohibitively high for mass-market hardware, consumer electronics brands are prioritizing native, local chip-level integration. By anchoring early-stage funding from state capital and primary manufacturing conglomerates, hardware-optimized software developers are positioning themselves to dictate the underlying technical standards for the next generation of intelligent devices.