Approaching.AI Joins MoMA to Upgrade the Compute-Network Eco
Recently, the 2026 Mobile Network Conference, held under the theme of co-creating a new form of the intelligent economy, opened in Guangzhou. During the conference, Approaching.AI (趋境科技), a high-quality AI token production service provider, formally signed an agreement to join China Mobile's MoMA Open Program. The company also won first prize in the Intelligent Computing Industry Innovation track of the 2026 "Mobile Cloud Cup" AI Application Innovation Competition, run alongside the conference, on the strength of its solid underlying systems engineering capabilities and proven real-world delivery results.

As a company deeply rooted in AI infrastructure, Approaching.AI had previously released the High-Quality AI Token Factory Solution for Domestic Heterogeneous Inference. Built on its industry-first full-system heterogeneous collaborative inference technology, the solution steadily converts diverse heterogeneous compute into cost-effective, production-grade AI token production capacity, forming a replicable system-level engineering path that supports carriers as they upgrade their compute-network ecosystems from delivering compute resources to delivering high-quality token services. This agreement and the award are a concrete demonstration of the company's system-level engineering capabilities within that ecosystem at the current stage.
At the technical level, Approaching.AI achieves precise coordination of heterogeneous resources with its full-system heterogeneous collaborative inference technology: resources with greater compute capability carry the high-compute-density Prefill phase, resources with larger memory bandwidth handle the Decode phase, and the CPUs, GPUs, memory, and high-speed networking in the cluster are brought into inference scheduling as a whole. Building on end-to-end preservation of model accuracy, fine-grained cross-GPU collaborative networking, and a millisecond-level fault-tolerance mechanism with zero data loss, the company enables resources with different compute and memory-bandwidth characteristics to participate collaboratively in inference, with each playing to its strengths, and measures optimization results by effective token output per unit of resource and effective production time. In practice, the solution has entered production, where it stably supports trillion-parameter ultra-large models and continuously handles high-concurrency real-world traffic. Measured results show that the overall price-performance ratio of the solution's output is better than that of advanced-compute baseline solutions, and that the same hardware spend can support larger volumes of high-quality AI token demand.
At the ecosystem level, the agreement brings Approaching.AI's high-quality AI token production and inference-system engineering capabilities into China Mobile's MoMA Open Program ecosystem. To connect token supply with application consumption, the company uses the Approaching.AI Token as a Service (ATaaS) platform to link underlying resources, model inference systems, and enterprise business scenarios, providing production-grade inference services that bring model capabilities into AI agents, intelligent hardware, and industry applications. To address the performance and cost challenges of scaled adoption, the company follows a simple principle: "Fewer Models. Deeper Optimization. Higher Quality." It focuses on a small number of highly productive models and carries out deep adaptation and system optimization against real business workloads, turning different types of compute resources into sustainably deliverable high-quality AI tokens. This optimizes unit cost while meeting business service requirements.
Looking ahead, Approaching.AI will continue to refine its full-stack system capabilities for the efficient use of diverse heterogeneous compute. It will put "Compute–Intelligence Integration, Shared Token Value" into practice through ecosystem collaboration and production delivery, and will serve the intelligent transformation of industry with a stable supply of high-quality AI tokens.
-本文版权归原作者所有,登载此文出于传递更多信息之目的,并不意味着赞同其观点或证实其描述。文章内容仅供参考,如有侵犯版权请来信告知E-mail:1308654573@qq.com,我们将立即处理。


