Application for Filecoin ProPGF Batch 3

Open Model

https://github.com/6block/openmodel

Gao Jiatong PR @carol9702

OpenModel is an open-source decentralized AI inference network built within the Filecoin ecosystem, providing a unified AI infrastructure layer for developers, applications, and compute providers. By aggregating distributed computing resources, OpenModel transforms idle compute capacity into accessible AI inference services, lowering the barrier for developers to integrate AI capabilities into their products. The core infrastructure maintained by OpenModel includes an Inference Gateway, model routing system, task scheduling system, developer APIs, and a distributed compute coordination mechanism. During this grant period, OpenModel will further expand its capabilities to support image generation, image understanding (via models such as Qwen-VL), and speech-related services, while building a unified multimodal AI service gateway on top of its existing scheduling framework. These infrastructure components serve third-party developers, AI applications, Filecoin ecosystem participants, and compute resource providers. Developers rely on OpenModel to rapidly access multimodal AI capabilities, applications depend on OpenModel for reliable inference services, and compute providers leverage OpenModel to monetize underutilized resources and unlock additional value from their infrastructure. OpenModel’s goal is to establish an open and sustainable AI infrastructure layer that enables developers and applications to build the next generation of AI products on top of the Filecoin ecosystem, while improving the utilization and economic value of distributed computing resources. Revenue model: Our inference gateway lets operators earn a share of the inference usage their users consume — the revenue that keeps our service running reliably for the long haul. Usage is metered per token and settled on-chain: a small operator fee, with the remainder paid directly to the storage providers whose GPUs served the request. Because the fee scales with usage rather than seats or subscriptions, operators earn more only as the network serves more — and every charge is independently verifiable against a provider-signed receipt.

Fully Open Source

OpenModel is committed to a fully open-source development model. Current open-source repositories include: * Node Infrastructure: https://github.com/6block/openmodel * Gateway Infrastructure: https://github.com/6block/openmodel-gateway The settlement and smart contract components are currently under development and will be released as open source upon completion.

During this grant period, OpenModel will focus on three key initiatives aimed at strengthening decentralized AI infrastructure and expanding ecosystem adoption within the Filecoin ecosystem. First, OpenModel will expand its multimodal capabilities beyond text generation by integrating image generation, image understanding (including models such as Qwen-VL), and speech-related services. These capabilities will be delivered through a unified API gateway built on OpenModel’s existing scheduling framework, providing developers with standardized access to multimodal AI services. Second, OpenModel will drive ecosystem growth and developer adoption. Through improved documentation, technical tutorials, community engagement, and developer support programs, OpenModel will demonstrate the value of utilizing idle compute resources for AI inference while attracting developers and ecosystem partners to build on the network. Third, OpenModel will develop demonstration applications, including AI assistants and content creation tools, to validate real-world use cases for decentralized AI infrastructure. These applications will showcase the capabilities of the network, generate meaningful AI workloads, and encourage broader adoption of decentralized AI services. Through these efforts, OpenModel will strengthen Filecoin’s AI infrastructure ecosystem, improve utilization of distributed compute resources, expand developer participation, and accelerate the adoption of open-source decentralized AI applications.

25,000

Indirect

Direct

Indirect

Output OpenModel will deliver four core outcomes during the grant period: * Build and maintain a fully open-source AI inference infrastructure and developer toolchain. * Expand multimodal AI capabilities, including image generation, image understanding, and speech-related services. * Improve developer documentation, SDKs, and ecosystem support resources. * Develop and launch demonstration applications, including an AI assistant and AI-powered content creation tools built on the OpenModel network. ↓ Outcome Developers will be able to access multimodal AI services through a unified API without the need to deploy and maintain complex AI models or inference infrastructure. More Filecoin ecosystem participants and third-party development teams will be able to build AI-powered applications on top of OpenModel, while additional compute providers can contribute resources to the network and participate in AI workloads. As developer adoption, application integrations, and inference requests increase, OpenModel will foster a more active AI ecosystem and improve the utilization of distributed compute resources. ↓ Impact OpenModel will enable more real-world AI workloads to run on open and decentralized infrastructure, creating a sustainable ecosystem that connects developers, applications, and compute providers. Through open-source infrastructure, multimodal AI capabilities, and real-world application validation, OpenModel will strengthen AI infrastructure within the Filecoin ecosystem, increase developer participation, and encourage broader adoption of decentralized AI services. Over time, this will help expand the utility, sustainability, and long-term value of the Filecoin ecosystem. As AI inference services continue to scale, OpenModel will create new economic value and revenue opportunities for the Filecoin ecosystem, further enhancing the network’s long-term economic sustainability.

Metric 1: Total Inference Requests Data Source: FVM Settlement Contract Event Logs Measurement: Total number of paid inference requests settled through the FVM settlement contract during the grant period. Each completed settlement represents a verified inference request processed by the OpenModel network. Target: 500,000+ requests Metric 2: Total Revenue Distributed to Storage Providers Data Source: FVM Settlement Contract State Measurement: Total amount settled and distributed from user deposits to participating Storage Providers through the settlement contract during the grant period. Target: USD 5,000+ equivalent Metric 3: Storage Providers Receiving Revenue Data Source: Storage Provider balances recorded in the FVM settlement contract Measurement: Number of Filecoin Storage Providers that have received positive settlement balances through the OpenModel settlement contract. Target: 5+ Storage Providers Metric 4: Active Developers Data Source: Developer API Key Usage Records Measurement: Number of unique API keys generating inference traffic during the grant period. Target: 50+ developers Metric 5: Third-Party Applications Integrated Data Source: Ecosystem integration records and API key attribution Measurement: Number of third-party applications built on OpenModel that generate real production traffic. Target: 5+ applications Metric 6: Production Models Available Data Source: OpenModel Model Registry Measurement: Number of production-ready AI models accessible through the OpenModel network. Target: 8+ models

OpenModel is currently in the infrastructure development phase and has not yet entered large-scale public adoption. As this grant progresses, we expect to onboard developers, ecosystem partners, Filecoin ecosystem participants, and compute providers to contribute to network growth and adoption, creating a broader base of stakeholders who can validate the project's impact. Potential references include Filecoin ecosystem partners, long-term Filfox ecosystem collaborators, developers integrating OpenModel APIs, compute providers participating in the network, and other projects and infrastructure partners contributing to the OpenModel ecosystem. Building on the team's long-standing experience operating Filecoin ecosystem infrastructure through projects such as Filfox and 6Block, OpenModel will continue to expand its network of ecosystem partnerships and foster a growing community of developers, users, and infrastructure participants.

The OpenModel team is composed of core members of the 6Block team, including infrastructure engineers, AI engineers, product leaders, and ecosystem operators with extensive experience in distributed systems, cloud computing, AI model services, and developer ecosystem growth. The team has long been involved in building decentralized infrastructure and has practical experience across the Filecoin ecosystem, blockchain systems, distributed compute networks, and AI application development. Team members have contributed to multiple open-source projects and infrastructure products, with experience spanning architecture design, network operations, product development, and ecosystem expansion. Building on the team's experience operating Filecoin ecosystem infrastructure such as Filfox and 6Block, OpenModel aims to establish an open, scalable, and sustainable decentralized AI inference network that provides reliable AI infrastructure for developers, applications, and compute providers.

OpenModel aims to serve as a bridge between decentralized infrastructure and the growing demand for AI applications. Through an open and fully open-source AI inference network, OpenModel enables developers to leverage distributed compute resources to build innovative AI-powered applications while advancing the adoption of AI within the Filecoin ecosystem. Beyond infrastructure development, the project also focuses on validating real-world use cases and application adoption. By supporting practical applications and real user demand, OpenModel seeks to demonstrate the long-term sustainability and value of decentralized AI networks while driving greater utilization of distributed compute resources.

The overall application process is well structured and provides a clear framework for projects to present their value proposition, implementation plans, and expected impact. For future funding rounds, it would be helpful to provide example applications for different project categories (such as infrastructure, developer tools, and application-layer projects), as well as more detailed KPI guidance and evaluation criteria. This would help applicants better understand reviewer expectations and improve the quality and consistency of submitted proposals.