ORA Protocol

创始人:

我们刚刚发布了 IQ AI.

查看详情

ORA Protocol

The Ora Protocol is a verifiable protocol that brings AI and complex computation on-chain. It was co-founded by in September 2022. [1][2]

Overview

Ora Protocol provides -agnostic infrastructure for integrating artificial intelligence with decentralized applications. It offers tools for developers to run and verify AI computations directly on-chain, aiming to create systems that are both trustless and decentralized. The protocol focuses on enabling scalable, verifiable AI within environments. Its primary service, the AI Oracle, allows to access AI functionalities through a decentralized network of using the TORA client. This system is supported by optimistic machine learning (opML) for ensuring computational integrity. Ora also introduces Initial Model Offerings (IMOs), a method for tokenizing AI models using ERC-7641. This approach supports funding for open-source AI development and distributes revenue to tokenholders tied to using these models in . [3]

Products

Onchain AI Oracle (OAO)

Ora Protocol’s on-chain AI Oracle is a verifiable oracle system designed to bring machine learning inferences onto the . It enables and to access and verify outputs from advanced AI models like LLaMA 3 and Stable Diffusion.

The system is powered by optimistic machine learning (opML), which uses fraud proofs to validate computations and ensure trustless execution. Compared to machine learning (zkML), this approach offers reduced computational cost and improved scalability.

The AI Oracle consists of and off-chain components. The opML contract manages fraud proofs and verifiability, while the AIOracle contract facilitates the interaction between off-chain ML nodes and on-chain callers. User contracts send AI requests and receive results. ORA Nodes, running the TORA client, are responsible for submitting and validating inference results across the network. [4]

Optimistic Machine Learning (opML)

opML (Optimistic Machine Learning) is a framework that enables scalable and efficient machine learning inference on networks. It allows advanced AI models to operate in decentralized environments without overloading the or relying on centralized systems. Inspired by , opML assumes ML computations are valid unless challenged, significantly reducing the need for immediate on-chain verification and making complex AI applications more practical on-chain.

The system achieves its efficiency through off-chain execution of ML tasks, with only disputed results undergoing on-chain validation. This is done via a fraud-proof mechanism: results are submitted by a service provider, and have a set period to challenge them. If a dispute occurs, the system identifies the exact computation step and verifies only that step using the Fraud Proof Virtual Machine (FPVM). This method preserves computational resources while ensuring accuracy and integrity.

Compared to Machine Learning (zkML), opML avoids the high costs of generating proofs, making it a more accessible solution for decentralized AI. It supports decentralization by keeping validation open and verifiable without centralized oversight. opML is available through ORA’s AI Oracle, providing developers with the infrastructure to integrate reliable and cost-effective AI features directly into and applications. [16]

Optimistic Privacy-Preserving AI

opp/ai (Optimistic Privacy-Preserving AI) is a framework designed to support both privacy and efficiency in on-chain machine learning by combining features from two existing approaches: Machine Learning (zkML) and Optimistic Machine Learning (opML). It allows machine learning models to be divided into components based on sensitivity, using zkML for computations that require data confidentiality and opML for less sensitive processes where speed and cost-efficiency are more critical.

The system works by partitioning a model into two parts. The first, M_zk, consists of submodels involving sensitive data or proprietary algorithms and is processed using zero-knowledge proofs to ensure this data remains private. The second, M_op, includes submodels where privacy is not a concern and is executed off-chain using opML’s optimistic approach, which assumes correctness unless a challenge is raised. This structure enables opp/ai to reduce the computational overhead typically associated with full zkML systems.

During execution, the outputs of the zkML submodels can be fed into the opML submodels, and vice versa, allowing for an integrated workflow. The zkML portion submits to the for verification, while the opML portion relies on a challenge-response model to confirm result accuracy. By blending these methods, opp/ai supports flexible, secure, and efficient deployment of AI models on networks. [17]

Resilient Model Services (RMS)

Resilient Model Services (RMS) is an AI infrastructure developed to support reliable and secure machine learning computations across various scenarios. Built on ORA’s opML (Optimistic Machine Learning) framework, RMS aims to ensure that AI processes remain stable, fault-tolerant, and verifiable. The initial stage of RMS includes ORA’s AI , which enables developers to access commonly used AI models for tasks like chat completion and image generation within a decentralized, verifiable environment.

RMS supports the development of autonomous that can operate transparently on-chain. These agents can execute tasks securely with on-chain verifiability, making them suitable for applications requiring high levels of trust and transparency. In addition, RMS facilitates the integration of AI with protocols, offering tools for implementing automated trading, risk assessment, and AI-based financial agents within on-chain systems.

gaming also benefits from RMS by enabling AI for responsive game mechanics, character interactions, and real-time decision-making. With -based verification of AI outcomes, RMS helps maintain fairness and transparency in gameplay. The foundation of RMS, opML, provides scalable, efficient, and decentralized AI computation while preserving transparency and trustworthiness across all supported use cases. [18]

Initial Model Offering (IMO)

The Initial Model Offering (IMO) is a mechanism designed to support the open-source development of AI models by tokenizing them on-chain. This approach aims to address the financial sustainability issues faced by developers in the open-source AI community, which currently operates under pressure from dominant, closed-source corporations. By turning AI models into tokenized assets, IMOs create a system where development and usage are more transparent, collaborative, and financially viable.

An IMO issues ERC-7641 tokens, called Intrinsic RevShare Tokens, representing ownership in an AI model's future value. These tokens entitle holders to a share of the usage fees generated when the AI model is used for inference or content generation. This structure aligns incentives between developers, users, and contributors, allowing each group to participate in the growth and governance of the model’s ecosystem.

The IMO process includes the tokenization of the model, followed by value creation through on-chain use, and distribution of resulting fees to token holders. Governance features allow token holders to influence the direction of the model's development, such as voting on grant allocations and future updates, reinforcing a community-driven approach to AI model evolution. [5] [6]

IMO Mechanism

The Initial Model Offering (IMO) mechanism is built around two core components: verifiable on-chain AI models and the ERC-7641 Intrinsic RevShare Token. This structure is designed to create a self-contained economic loop in which AI model usage on the directly generates and distributes value through a standardized token system.

  • Verifiable on-chain AI models are made possible through opML, a framework that applies fraud proofs to validate machine learning computations efficiently and scalable. These models are deployed through ORA’s AI Oracle, which serves as the infrastructure enabling AI inferences to be executed and verified on-chain. This setup ensures that each time a model’s output is used, such as by a , a verifiable interaction takes place that can trigger revenue capture.
  • The ERC-7641 token facilitates this revenue distribution by linking model usage to token-based rewards. A fee is collected and proportionally distributed among token holders whenever the AI model is used on the chain. Additionally, if the model produces content with royalties or fees (such as ), those revenues can be distributed using related standards like ERC-7007. This design supports sustainable development and decentralized participation in the AI model’s economic ecosystem. [5]

ERC-7641

ERC-7641 is an extension of the token standard. It integrates a built-in revenue-sharing mechanism, enabling holders to claim a proportional share of a communal revenue pool based on their token balance at designated snapshot intervals. It supports sustainable funding by allowing projects to tokenize revenue streams and distribute income fairly among participants. The standard also includes a function, where holders can destroy their tokens in exchange for a share of the pool’s value, introducing deflationary economics that reduce supply and incentivize long-term participation. [19]

ERC-7007

ERC-7007 is an extension of the standard designed to represent and verify AI-generated content on the . It enables to authenticate AI-generated content, ensuring it is produced by a specific model using a given input. The standard integrates Machine Learning (zkML) and Optimistic Machine Learning (opML) to verify content correctness. By providing a way to verify and monetize AI-generated content, ERC-7007 supports AI model authors and content creators, offering a secure and standardized method to facilitate revenue-sharing through verifiable . It includes components such as the AI model, zkML/opML verification, and a compliant AIGC- for managing content and verification processes. [20]

ORA

The $ORA token is integral to the ORA ecosystem, providing multiple functions. It grants access to Initial Model Offerings (IMOs), enabling members to fund and support the development of cutting-edge open-source AI models. Token holders can $ORA to operate decentralized , which support services like OAO and RMS while earning . $ORA also facilitates decentralized governance, allowing holders to influence the protocol's direction. Additionally, $ORA serves as universal , enabling -free interactions across chains and protocols by converting $ORA into the native required for transactions. [21]

Tokenomics

ORA has a total supply of 333,333,333 tokens and has the following distribution: [21]

  • Ecosystem: 33%
  • Foundation: 33%
  • Network: 33%
  • DAICO: 1%

Team

  • - Co-founder & Generator [9]
  • - Ecosystem Growth [10]
  • Cathie So - Chief Scientist [11]
  • Shuxiao Miao - Software Developer [13]
  • Suning Yao - Research Engineer [12]
  • Levi Sledd - Zero Knowledge Circuit Engineer[14]

Partnerships

  • MemeWar
  • Taste Foundation
  • Hetu Protocol

价格

$0.848207

1.53%

市值

$32,516,715.00

1.19%

稀释市值

$280,878,194.00

1.19%

成交量

$329,896.60

0.16%

ORA

ORA

USD

USD

编辑者

Profile picture of Anonymous userSophIA

编辑日期

May 1, 2025

编辑原因:

Republishing the ORA Protocol wiki with updated content and links.

Loading...

参考文献

首页分类排名事件词汇表