Preface

The large-scale adoption of large models and intelligent agents is propelling the intelligent economy into a new stage of development. Token has evolved from a basic technical unit of measurement into a new factor of production that connects energy, chips, datasets, and large models. The Token economy has officially become the core theme of industrial transformation.

At the policy level, the 2026 Government Work Report explicitly calls for developing the intelligent economy, and the National Data Administration is advancing the marketization of Token. On the market side, domestic daily Token calls have surpassed 140 trillion, achieving a thousand-fold growth in two years—a clear trend of industrial change.

The so-called Token economy is a new industrial operation model that uses Token as a unified carrier for measurement, pricing, and value assessment, linking computing supply, full-domain operations, and scenario implementation. Most enterprises have not yet established a matching Token management mechanism, making it difficult to accurately quantify the value of intelligent capabilities—a standardized operation system has become a common industry need.

Authoritative Insights

CAICT Report Explains the Logic of the Token Economy; BONC Deeply Involved in Compilation

The “Token-Driven Intelligent Economy Research Report (2026)” led by the China Academy of Information and Communications Technology (CAICT) and jointly compiled by nearly a hundred research institutions and industry players has been officially released. Dr. Zha Li, Vice President and CTO of BONC, participated as a core compiler throughout the report’s discussions, trend analysis, and viewpoints, integrating frontline enterprise practice into the research on the development path of the Token economy. The report is a systematic research achievement of significant reference value in China’s Token economy field, comprehensively outlining the evolution of Token, the three-tier industrial ecosystem, and the challenges of large-scale implementation.

01 Token Completes a Three-Level Value Leap, Laying the Factor Foundation for the Token Economy

The report outlines Token’s complete evolution path: in its early days, Token served mainly as a text-splitting unit in NLP; after the Transformer architecture became widespread, Token became the core computing unit throughout the encoding, inference, and generation of large models; today, Token further integrates underlying resources such as energy, chips, high-quality datasets, and large models, gradually evolving into a standardized industrial factor that is measurable, priceable, circulable, and traceable. With the continued marketization of data factors, building a standardized Token operation system has become an important foundation for enterprises to deploy AI at scale.

02 Three-Tier Collaborative Architecture, Building the Complete Industrial Framework of the Token Economy

The report predicts that the Token economy will form a three-tier closed-loop industrial ecosystem of “production & supply – scheduling & distribution – end consumption”: the bottom-level Token factory completes standardized Token production; the middle-level scheduling and distribution hub coordinates measurement, pricing, resource allocation, and circulation services; the top level implements intelligent applications for various business scenarios, continuously releasing business value. To implement the Token economy, enterprises need to benchmark the three-tier architecture and build a complete system covering output, control, and application, achieving measurable, priceable, and effective Tokens.

03 Three Major Management Shortcomings in the Industry Constrain the Large-Scale Adoption of the Token Economy

The report summarizes common challenges in current industrial implementation: first, industry standards such as Token measurement calibers and interaction interfaces are not yet fully unified, limiting cross-platform collaboration efficiency; second, enterprises manage computing and model resources in a fragmented way, making it difficult to accurately account for Token usage data by department, project, or business line, and hard to quantify AI ROI; third, the industry chain lacks mature cross-entity circulation, market-based pricing, and automated profit-sharing mechanisms, keeping Token collaboration and circulation costs high.

Landing Practice

BONC Builds a Full-Chain Integrated Token Economy Operation Solution

With years of deep cultivation in industrial digitalization, relying on its own computing layout, industry model service capabilities, and enterprise service experience, BONC has completed a strategic transformation from a traditional software service provider to an AI refined operation service provider. Against the report’s three-tier industrial architecture, BONC has built a complete enterprise Token operation system covering computing supply, operation control, and business implementation, systematically addressing the three core challenges enterprises face in large-scale AI implementation: stable supply, cost transparency, and value realization.

01 Token Supply Side: Building a Stable, High-Quality, Sustainable Token Capacity System

The primary challenge for enterprises implementing the Token economy is capacity stability and economy. BONC systematically addresses this from three dimensions: computing scheduling, inference enhancement, and access standardization.

1.1 Heterogeneous Computing Unified Pooling

Compatible with multiple categories of domestic and foreign computing hardware, through heterogeneous computing unified pooling and scheduling, it achieves integrated management of computing, storage, and network resources, supporting cluster elastic expansion, resource isolation, and full-link cost accounting, optimizing hardware utilization to the maximum.

1.2 Inference Enhancement Unleashes Computing Efficiency

Building a CPU+GPU+memory+SSD collaborative inference enhancement system, through key technologies such as heterogeneous hybrid inference, storage-computing collaborative optimization, model state hierarchical caching, KV Cache reuse, and resource virtualization scheduling, it optimizes the computing, storage, and scheduling efficiency in the model inference process, lowering the deployment threshold and unit Token production cost, and improving the responsiveness, stability, and overall computing efficiency of large model services.

1.3 Multi-Model Standardized Unified Access

Encapsulating various models into standardized general protocols, enterprises only need to maintain one set of business code to flexibly switch underlying models, greatly reducing the labor and technical costs of parallel multi-model O&M, and building a low-cost, high-efficiency, sustainably iterable stable Token supply system.

02 Token Operation Side: From Cost Black Box to Refined Measurement and Value Operation

For the Token economy to truly take effect, enterprises must establish comprehensive perception and control capabilities over Token consumption. BONC has built a refined operation system that is measurable, priceable, and effective, covering the complete control chain from perception to decision-making.

2.1 Multi-Dimensional Full-Volume Consumption Measurement, Building a Visual Cost Ledger

Supporting the aggregation of Token usage data by multiple dimensions such as department, project, employee, and business application, with a real-time operation dashboard displaying call scale, model distribution, and interface running status, transforming originally opaque resource consumption into traceable operation data.

2.1 Tiered Quota Precise Control, Building a Full-Link Compliance Baseline

Through flexible configuration of Token usage quotas, validity periods, and accessible model scopes for each business entity, combined with sub-token distribution, IP whitelist, and full-process audit logs, it meets security compliance requirements while ensuring flexible resource allocation.

2.3 Differentiated Dynamic Pricing, Upgrading Value-Added Services

Abandoning fixed packaged plans, relying on real operation data to customize differentiated Token service plans for enterprises, adding semantic caching mechanisms to reduce overall usage costs, and outputting AI efficiency diagnosis and resource allocation optimization suggestions, upgrading the service connotation from Token supply to full-lifecycle value-added operations.

03 Token Application Side: Business Tokenization Implementation, Building a Quantifiable Value Loop

3.1 Ontology-Driven, Full-Scenario Business Tokenization Implementation

Relying on the Agentic Ontology platform to build a unified business super-entry, Token capabilities are deeply embedded into core business processes such as R&D, finance, legal, customer service, and business analysis, deeply integrated with actual business scenarios, forming a continuous consumption mechanism.

3.2 Closed-Loop Mechanism, Quantifiable AI ROI

The entire system builds a positive-growth closed loop of “business requirement implementation – Token call statistics – business value assessment – model iteration optimization,” where every Token consumption corresponds to specific business output, making AI investment effects quantifiable and reviewable, continuously optimizing enterprise resource allocation strategies.

3.3 Practice Validation, High-Value Scenarios Continuously Releasing Efficiency

Currently, BONC has achieved full-staff coverage of AI capabilities internally, with Token calls mainly focused on high-frequency business scenarios such as software R&D, data analysis, contract review, document writing, process automation, and customer operation, promoting AI’s gradual integration from single-point tool applications into daily business processes.

On this basis, BONC further explores a new paradigm of Token-driven intelligent applications with “Skill+Ontology” as the core path: through ontology modeling, business objects, relationship rules, scenario knowledge, and agent capabilities are structurally organized, so that Token calls no longer stay at simple Q&A or scattered calls, but are deeply embedded in business scenarios, supporting task execution, semantic reasoning, and value transformation. For typical business scenarios, ontology capabilities have formed applications in directions such as cultural resource digitalization and customer operation: in cultural resource digitalization, it enables systematic integration, semantic expression, structured organization, and deep mining of historical materials; in customer operation, it reconstructs the business semantic relationships among customers, products, touchpoints, and strategies, improving renewal-intent identification, risk stratification, strategy matching, and intelligent planning capabilities. From the perspective of landing results, Token is evolving from a model-calling resource into an important carrier driving business application implementation and value creation.

Standardized Token supply, refined full-domain operations, and scenario-based value landing form a complete closed loop, providing a systematic practice path for enterprises to build a sustainably operating Token economy mechanism.

Industry Recognition

Landing Practice Recognized by Industry and Academia; BONC Shares Token Economy Solution at CCF Conference

Relying on mature enterprise-level Token economy full-chain landing practice, Ma Hongwei, General Manager of BONC’s Agent System Department, was invited to attend the first CCF Large Model and AI Engineering Conference held in Tianjin from June 11 to 13, delivering a keynote speech titled “From Cost Black Box to Refined Operations: MeshToken’s Practice in Building Enterprise Token Economy.”

The conference gathered more than 400 university scholars, research experts, and industry leading enterprise representatives, focusing on deep exchanges around large model engineering and intelligent industry ecosystems. The speech comprehensively dissected the logic of building an enterprise Token economy, the full-chain technical architecture, and real landing cases, outputting replicable engineering solutions for core pain points such as industry computing coordination and Token value quantification, triggering widespread industry discussion and recognition.

From deeply participating in the compilation of the Token industry report to outputting practical templates at academic-industry summits, it fully proves that this full-chain operation system has industry-wide adaptability and mature landing capabilities.

Ecosystem Win-Win

Building an Open and Win-Win Industrial Ecosystem, Broadening the Boundaries of Token Economy Development

For the long-term development of the Token economy, BONC will continuously iterate Token full-lifecycle operation capabilities, deeply participate in the co-construction of Token industry standards; link up computing, model, and application industry chain partners to open up Token cross-entity circulation and standardized settlement channels, building an inclusive and win-win Token industry ecosystem. At the same time, it will continue to deepen the real economy tracks such as manufacturing, finance, government, energy, and medicine, output standardized enterprise Token economy operation plans, and help all walks of life maximize the value of intelligent capabilities relying on the standardized Token system.

Closing

Token empowers thousands of industries; refined operations enlighten the future. The era of the intelligent economy offers broad opportunities. BONC looks forward to joining peers from all walks of life, relying on the guidance of the industry report, and using the maturely landed enterprise AI full-chain Token operation solution as the starting point, to jointly explore the industrial value of the Token economy and inject continuous momentum into the high-quality development of the domestic intelligent economy.