OpenLedger Depth Research Report: Building a data-driven, model-composable agent economy based on OP Stack + EigenDA

1. Introduction | The Model Layer Leap of Crypto AI

Data, models, and computing power are the three core elements of AI infrastructure, and none can be omitted. The Crypto AI field has undergone an evolution path similar to that of the traditional AI industry. At the beginning of 2024, the market's focus is on decentralized GPU projects, emphasizing competition in computing power. As we enter 2025, the industry's focus gradually shifts towards the model and data layers, marking a transition for Crypto AI from competition over underlying resources to a more sustainable and application-value-driven mid-layer construction.

General Large Model (LLM) vs Specialized Model (SLM)

Traditional large language model (LLM) training relies on large-scale datasets and complex architectures, with a huge parameter scale and high training costs. Specialized language models (SLM), on the other hand, are based on open-source models, combining a small amount of high-quality professional data and technologies such as LoRA to quickly build expert models in specific fields, significantly reducing training costs and technical barriers.

SLM collaborates with LLM through Agent architecture calls, plugin system routing, LoRA module hot swapping, RAG, and other methods, retaining the wide coverage capabilities of LLM while enhancing professional performance through fine-tuning modules, forming a flexible combinatorial intelligent system.

The value and boundaries of Crypto AI at the model layer

Crypto AI projects find it difficult to directly enhance the core capabilities of LLM, largely due to technical barriers and limitations of the open-source ecosystem. However, on top of open-source foundational models, Crypto AI projects can achieve value extension by fine-tuning SLM and integrating the verifiability and incentive mechanisms of Web3. Its core value is reflected in two directions: the trust verification layer and the incentive mechanism.

AI Model Type Classification and Blockchain Applicability Analysis

The model-based Crypto AI projects mainly focus on small SLM fine-tuning, on-chain data access and verification of RAG architecture, as well as local deployment and incentives for Edge models. Combining the characteristics of blockchain, Crypto can provide unique value for these low to medium resource model scenarios, forming differentiated value for the AI "interface layer."

The blockchain AI chain based on data and models can provide clear and immutable on-chain records of the sources of data and model contributions, enhancing credibility and traceability. Through the smart contract mechanism, reward distribution is automatically triggered when data or models are invoked, transforming AI behavior into measurable and tradable tokenized value. Community users can also participate in governance through token voting, improving the decentralized governance structure.

2. Project Overview | OpenLedger's AI Chain Vision

OpenLedger is a blockchain AI project focused on data and model incentive mechanisms. It proposes the concept of "Payable AI" aimed at building a fair, transparent, and composable AI operating environment that incentivizes all participants to collaborate and earn on-chain rewards.

OpenLedger provides a complete closed-loop from "data provision" to "model deployment" to "revenue sharing", with core modules including Model Factory, OpenLoRA, PoA, Datanets, and the model proposal platform. Through these modules, OpenLedger has built a data-driven, model-composable "intelligent economy infrastructure."

On blockchain technology, OpenLedger adopts OP Stack + EigenDA as its foundation, providing a high-performance, low-cost, and verifiable operating environment for AI models. It is based on the Optimism tech stack, supporting high throughput and low fees; it settles on the Ethereum mainnet; EVM compatible; EigenDA provides data availability support.

Compared to the more underlying NEAR, OpenLedger focuses more on building an AI-specific chain oriented towards data and model incentives, aiming to achieve traceable, composable, and sustainable value loops for model development and invocation on the chain.

Three, the Core Components and Technical Architecture of OpenLedger

3.1 Model Factory, no-code model factory

ModelFactory is an LLM fine-tuning platform under the OpenLedger ecosystem, providing pure graphical interface operations. Its core processes include data access control, model selection and configuration, lightweight fine-tuning, model evaluation and deployment, interactive verification interface, and RAG generation tracing.

ModelFactory supports mainstream open-source large language models, such as the LLaMA series, Mistral, Qwen, ChatGLM, Deepseek, Gemma, etc. Although it does not include the latest MoE or multimodal models, it has made a "practical first" configuration based on the realistic constraints of on-chain deployment.

Model Factory, as a no-code toolchain, has all models built-in with a contribution proof mechanism to ensure the rights of participants, featuring advantages of low barriers to entry, monetizability, and composability.

3.2 OpenLoRA, the on-chain assetization of fine-tuned models

OpenLoRA is a lightweight inference framework built by OpenLedger, designed to address issues such as high costs, low reusability, and resource waste in AI model deployment. Its core components include the LoRA Adapter storage module, model hosting and dynamic fusion layer, inference engine, request routing, and streaming output module.

OpenLoRA significantly enhances multi-model deployment and inference efficiency through a series of underlying optimizations. Its core includes dynamic LoRA adapter loading, tensor parallelism, Paged Attention, multi-model fusion, Flash Attention, precompiled CUDA kernels, and quantization techniques.

OpenLoRA is not only an efficient inference framework, but it also deeply integrates model inference with Web3 incentive mechanisms, aiming to turn LoRA models into callable, composable, and revenue-sharing Web3 assets. It supports model-as-asset, dynamic merging of multiple LoRAs + revenue attribution, and features such as multi-tenant shared inference for long-tail models.

3.3 Datanets, from data sovereignty to data intelligence

Datanets is the infrastructure of OpenLedger "Data as Assets" for collecting and managing datasets in specific fields. Each Datanet acts like a structured data warehouse, ensuring data traceability and trustworthiness through an on-chain ownership mechanism.

Compared to projects focusing on data sovereignty, OpenLedger builds a complete closed loop of "from data to intelligence" through three modules: Datanets, Model Factory, and OpenLoRA, focusing on how data is trained, invoked, and rewarded.

3.4 Proof of Attribution: Restructuring the Incentive Layer for Benefit Distribution

PoA is the core mechanism of OpenLedger for data ownership and incentive distribution. Its process includes data submission, impact assessment, training validation, incentive distribution, and quality governance. PoA is not only a tool for incentive distribution but also a framework aimed at transparency, source tracing, and multi-stage ownership.

RAG Attribution is the data attribution and incentive mechanism established by OpenLedger in the RAG scenario, ensuring that the model output is traceable and verifiable, contributors can be incentivized, and ultimately achieving credibility in generation and transparency in data.

4. OpenLedger Project Progress and Ecological Cooperation

OpenLedger has launched its testnet, with the data intelligence layer as the first phase, aimed at building a community-driven internet data warehouse. The testnet offers three types of reward mechanisms: node operation rewards, data contribution rewards, and task participation rewards.

Epoch 2 testnet focuses on the launch of the Datanets data network mechanism, covering tasks such as data validation and classification. OpenLedger's long-term roadmap plans to move from data collection and model building to the Agent ecosystem, gradually achieving a complete decentralized AI economic closed loop.

OpenLedger's ecosystem partners encompass computing power, infrastructure, toolchains, and AI applications. Over the past year, OpenLedger has continuously hosted the DeAI Summit, strengthening its brand recognition and professional reputation within the developer community and the Web3 AI entrepreneurial ecosystem.

5. Financing and Team Background

OpenLedger completed a $11.2 million seed round of financing in July 2024, with investors including well-known institutions such as Polychain Capital, Borderless Capital, and several angel investors. The funds will be used to advance the construction of the AI Chain network, model incentive mechanisms, data infrastructure, and the comprehensive implementation of the Agent application ecosystem.

OpenLedger was founded by Ram Kumar, an experienced entrepreneur in the fields of AI/ML and blockchain technology, who brings an organic combination of market insight, technical expertise, and strategic leadership to the project.

6. Token Economic Model Design and Governance

OPEN is the core functional token of the OpenLedger ecosystem, empowering network governance, transaction operation, incentive distribution, and AI Agent operation. Its functions include governance and decision-making, transaction fuel and fee payment, incentives and belonging rewards, cross-chain bridging capabilities, and AI Agent staking mechanisms.

OpenLedger introduces a governance mechanism based on contribution value, where voting weight is related to the actual value created, rather than purely capital weight. This design helps to achieve long-term sustainability in governance and prevents speculation from dominating decision-making.

7. Data, Models, and Incentive Market Landscape and Competitor Comparison

OpenLedger occupies a middle layer position in the current Crypto AI ecosystem, serving as a key bridging protocol that connects model value supply with practical application incentives. Compared to other projects:

• Protocol Incentive Layer: OpenLedger vs. Bittensor
• Model Ownership and Call Incentives: OpenLedger vs. Sentient
• Model Hosting and Trustworthy Inference Platform: OpenLedger vs. OpenGradient
• Crowdsourcing Model and Evaluation Incentives: OpenLedger vs. CrunchDAO
• Community-driven lightweight model platform: OpenLedger vs. Assisterr
• Model Factory: OpenLedger vs. Pond
• Trustworthy Inference Path: OpenLedger vs. Bagel
• Data collaboration path: OpenLedger vs. Sapien / FractionAI / Vana / Irys

8. Conclusion | From Data to Model, the Monetization Path of AI Chain

OpenLedger is committed to building the "model as an asset" infrastructure in the Web3 world, bringing AI models into a truly traceable, monetizable, and collaborative economic system for the first time through the construction of a complete closed loop. Its technical system provides comprehensive support for all participants, activating the long-ignored resources at both ends of the AI value chain: "data" and "models."

OpenLedger is more like HuggingFace + Stripe + Infura in