Monai is a decentralised artificial intelligence (AI) project focused on building censorship-resistant large language models (LLMs). It is designed to allow open access to generative AI tools that operate independently of corporate or governmental control. The project's foundational principle is that AI should be free from moderation filters, biases, or centralised oversight.

At its core, Monai aims to create an economically viable ecosystem for decentralised LLMs, beginning with Merovingian I—a multilingual model trained on trillions of tokens across varied domains such as Reddit, Wikipedia, PubMed, and more. Monai is being integrated with Monad, a high-throughput blockchain, to support decentralised compute infrastructure and enable efficient inference processing.

The platform uses a Transformer-based architecture, similar to other leading LLMs, and will progressively include multi-modal features such as text-to-speech and image generation. Unlike centralised models, Monai does not implement hard-coded biases or moderation layers. Its training datasets are designed to fill knowledge gaps found in mainstream models, and its models will be frequently retrained with new data and community feedback.

The Monai protocol also introduces a dynamic query routing system that allocates inference requests to compute nodes based on uptime, latency, and tokens staked. This incentivises reliable and performant participation, while also supporting the decentralisation of AI infrastructure.

The MONAI token is the native utility asset of the Monai ecosystem. It plays a central role in facilitating the economic and computational interactions that support the protocol:

• Access to Inference: Users pay MONAI to query the AI models. These queries can be one-off or through subscription-based models.
• Incentives for Compute Providers: GPU node operators earn MONAI by providing computing resources needed to process model inferences.
• Staking and Delegation: Token holders can stake or delegate MONAI to node operators, influencing which nodes receive more queries and rewards. This staking also subjects node operators to slashing if they perform poorly or submit wasteful governance proposals.
• Governance Participation: Governance proposals can include adopting new models, hardware upgrades, or identifying issues like censorship. Proposers are incentivised through rewards from a governance treasury initially funded with 2.5% of network rewards.

The token was launched on Ethereum to enable broader accessibility. A 1:1 migration to Monad is planned once the Monad mainnet is operational.

The Monai project is led by a multidisciplinary team with expertise in AI, blockchain, and systems development:

• Jarvis (Business Development Lead) – Previously worked in AI since 2017 and has a background in venture capital.
• Dr. Henry (Chief Technology Officer) – Holds a Ph.D. in Computer Science with a specialisation in neural networks and over 15 years of experience in AI R&D.
• Josh (Blockchain Lead) – Focused on decentralised system integration and the Monai-Blockchain interface.
• Jordan (Systems Developer) – Ensures non-AI infrastructure runs efficiently.

The machine learning team includes engineers, data scientists, and researchers with prior experience in major AI labs and projects. Their collective responsibilities include refining Monai's training datasets, enhancing model architecture, and ensuring continuous performance improvements.

The team credits foundational research by OpenAI, Google, and others as vital for making advanced models possible, and draws on academic literature for moderation, safety, and bias-related challenges.