--- title: "ModelWare and China Telecom pair edge-side AI with a national network" date: 2026-10-31 category: Companies & Stack site: NeuroAI canonical: https://neuroai.site/a/na-corp-modelbest-chinatelecom-partnership language: en --- # ModelWare and China Telecom pair edge-side AI with a national network > On Oct 12, 2026, edge-side model specialist ModelWare and China Telecom's AI arm signed a strategic pact to push on-device AI into phones, PCs, and industry. ## Key takeaways - On October 12, 2026, ModelWare (面壁智能) and China Telecom's AI unit signed a strategic cooperation agreement in Beijing. - ModelWare brings edge-side (端侧) base models — its MiniCPM series — already running in cars, smart homes, AI phones, and AI PCs. - China Telecom contributes cloud-side models, a nationwide compute-and-channel network, and the Tianyi AI (天翼AI) edge-cloud product system. - The pact targets two fronts: consumer-edge AI products and enterprise vertical models plus agent (智能体) infrastructure. - Both sides will jointly build solutions, jointly go to market, and co-develop inference-acceleration technology. - The thesis: "cloud training, edge deployment, edge-cloud collaboration (端云协同)" is how AI reaches billions of terminals. ## Two companies, one gap On October 12, 2026, in Beijing, a small but symbolically neat deal was signed. ModelWare (面壁智能), a Beijing-based specialist in edge-side (端侧) AI, and China Telecom's artificial-intelligence arm, China Telecom AI (中电信人工智能科技), put their names on a strategic cooperation agreement. Lei Shengtao, ModelWare's president, and Yang Ge, deputy general manager of China Telecom AI, signed for their sides. The logic is straightforward. ModelWare is strong where the network ends — on the device. Its MiniCPM series of edge-side (端侧) base models (基座大模型) already power scenarios in automobiles, smart-home gear, AI phones, and AI PCs. China Telecom owns the opposite strength: cloud-side models, a cloud-network integrated infrastructure, and sales channels that reach essentially every province. Neither alone closes the loop; together they claim to. ## What the deal actually covers According to both companies, the cooperation runs on a "joint solutions, joint promotion" model across two directions. On the consumer side, the aim is personalized edge-side (端侧) AI applications for phones, PCs, and smart homes. On the enterprise side, the focus is on standardized vertical-industry models and agent (智能体) infrastructure for a broader set of customers. Beyond shipping products, the two said they will keep improving edge-side (端侧) model capability and jointly explore inference-acceleration technology — the unglamorous engineering that decides whether a model on a phone feels instant or sluggish. The stated end state is a closed loop of "cloud training, edge deployment, edge-cloud collaboration (端云协同)," with ModelWare's models folded into China Telecom's Tianyi AI (天翼AI) edge-cloud system. ## Why edge-side, why now ModelWare's bet is captured in what it calls a "density law": knowledge density keeps rising while the cost of intelligence keeps falling, so the value of a model is finally realized on every terminal, not just in a data center. Edge-side (端侧) models are what turn a phone or a PC into a physical-AI entry point. That is a different framing from the "bigger is better" race and fits a carrier that serves hundreds of millions of devices. The partnership also lands at a moment when on-device AI is moving from demo to default. Just a day before the signing, on October 13, 2026, ModelWare and the OpenBMB community open-sourced MiniCPM-V 4.7, a multimodal edge model in 35B-A3B and 1B versions — evidence the company is shipping the small, capable models the deal needs. ## What each side gets For ModelWare, China Telecom is a distribution rocket. Building a great edge-side (端侧) model is hard; getting it onto tens of millions of terminals at carrier scale is the harder part, and that is exactly what a national operator provides. For China Telecom, ModelWare supplies a differentiated edge-side (端侧) capability to plug into its cloud-AI story, at a time when every carrier wants to show it can deliver AI, not just connectivity. Crucially, both sides framed the signing as a step up from "project-level exploration" to "strategic cooperation" — the kind of language that signals a multi-year commitment rather than a one-off pilot. ## The bigger pattern This is not an isolated handshake. China's carriers have spent two years repositioning from connectivity utilities into AI distributors, and on-device intelligence is the piece that lets them reach users without sending every query to a distant data center. Pairing a specialist edge-side (端侧) model maker with a national operator mirrors how the US market links chip and device makers to carriers and cloud providers — except here the bundling is explicit and state-backed at the distribution layer. If the "cloud training, edge deployment" loop proves reliable, expect more such pairings across appliances, vehicles, and industrial gear. ## Honest limitations Core facts (signing date, signatories, scope, the two cooperation directions, the edge-cloud framing, the MiniCPM-V 4.7 release) are drawn from ModelWare's official site and corroborated by Chinese financial and technology press (China Economic News, Sina, 10jqka). We did not see a disclosed deal value, so none is reported. The strategic claims about market impact are the companies' own and are not independently verified here. Whether the "cloud training, edge deployment" loop produces shipping, revenue-bearing products remains to be seen; carrier-AI partnerships have a long history of announcements outrunning deployments. ## What readers can do now - If you build consumer hardware, watch ModelWare's MiniCPM releases as a low-cost way to add on-device intelligence without a full in-house model team. - Enterprises evaluating carrier AI should ask China Telecom specifically about edge-cloud (端云协同) reference deployments, not just cloud API access. - Treat this as a signal: edge-side (端侧) AI plus national distribution is becoming a real go-to-market pattern in China, worth tracking against pure-cloud offerings. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-corp-modelbest-chinatelecom-partnership Free to quote with attribution and a link to the original.