Model Marketplace & Curation

The InfraMind Model Marketplace is the open registry and discovery layer for deployed models on the mesh. It serves as both an execution directory and a decentralized exchange of AI capabilities—where developers can list models, users can access them permissionlessly, and the DAO can curate, incentivize, or govern the lifecycle of deployed intelligence.

Unlike traditional AI APIs where hosting and access are controlled by a centralized gatekeeper, InfraMind’s marketplace is fully decentralized: models are containerized, uploaded to distributed storage (IPFS/Arweave), and referenced through verifiable content hashes. Access, pricing, discoverability, and even model deprecation are governed by protocol rules and community input.

Listing Models for Public Use

Any developer can deploy and register a model using the CLI:

infra register --model model.yaml --name summarizer-v1 --version 1.0

Registered models appear in the marketplace:

  • With public metadata (name, version, runtime, schema)

  • Linked to their storage hash (ipfs://...)

  • Marked with verification badges if schema-validated

  • Automatically assigned a global endpoint:

    https://api.inframind.host/inference/v1/summarizer-v1

Each listing includes:

  • A container reference hash

  • Verified JSON input/output schemas

  • Token pricing info (if applicable)

  • Runtime diagnostics (latency, availability, success rate)

  • Ownership metadata and reward routing address

All model endpoints are backed by mesh-executed nodes; no static server or API host is involved.

Curated Lists

To ensure quality and aid discoverability, the marketplace includes curated collections maintained by InfraDAO and trusted indexers.

Curated lists include:

  • Top Performing Models (ranked by usage and reputation)

  • DAO-Pinned Models (nodes are incentivized to cache these)

  • Privacy-Ready Models (compatible with enclave or zkML runtime)

  • Experimental Models (flagged as unstable or alpha)

  • Agent-Compatible Models (return structured output usable in chaining)

Curators can be DAO-elected or rotate via token-vote epochs.

Models that repeatedly underperform or exhibit schema mismatch may be delisted from featured views (though still accessible via direct call).

Curated models are listed at:

https://app.inframind.host/models/featured

Private Endpoints

InfraMind supports private model listings, which do not appear in the public registry and require authorization tokens to access.

In model.yaml:

access:
  public: false
  auth_required: true
  allow_list:
    - 0xB71c...     # wallet addresses
    - 0xF4c9...

Private endpoints are ideal for:

  • Commercial models (e.g. GPT-style or proprietary finetunes)

  • Enterprise pipelines with sensitive data

  • Whitelisted agent groups

  • Paywalled inference services

Authorized users can generate scoped tokens:

infra token generate --model agent-vault --expires-in 24h

These tokens are submitted in the Authorization: Bearer header.

Private models can still use the full mesh for execution—only the call is gated.

Model Bounties

The marketplace allows the DAO and community to post bounties for missing models or tooling that could enhance the mesh.

Example bounties:

Bounty
Reward (IFM)
Requirements

Finetuned Whisper (de)

3,000 IFM

German speech-to-text container (fastapi)

Vision Model (Quantized)

4,200 IFM

8-bit ResNet+ViT hybrid (≤500MB container)

WASM-Deployable T5

2,000 IFM

Stateless summarizer in Rust/WASI

LLM Profanity Filter

1,500 IFM

Schema validator + blacklist return model

Token Embedder (Multi-lang)

2,500 IFM

SentenceTransformer with REST IO schema

To submit for a bounty:

infra bounty submit --bounty-id 432 --model ./whisper-de

Submissions are evaluated by the DAO or bounty proposer. Once approved, rewards are unlocked to the deployer’s wallet.

Accepted bounty models are:

  • Listed under a dedicated bounty-complete tab

  • Pinned to nodes for cache preference

  • Rewarded with usage-based bonus multipliers

Community members can also propose new bounties via governance.

Model Lifecycle

Each model moves through several lifecycle stages:

  1. Pending – Registered but not validated or invoked

  2. Active – Passed schema checks, verified container

  3. Popular – High usage and uptime, pinned by default

  4. Deprecated – Replaced by newer version (soft redirect)

  5. Delisted – Flagged for failure, abuse, or inactivity

Model owners can deprecate manually:

infra model deprecate --name summarizer-v1

Or schedule auto-expiry in the manifest:

expires_at: 1720600000

Economics & Revenue

Model owners can set per-request fees:

pricing:
  per_request_fee: 0.25 IFM
  payout_address: 0xABC123...

They receive a share of execution fees after mesh rewards are distributed. If usage exceeds a protocol-defined threshold, models may be featured or auto-pinned with treasury subsidy.

Revenue is calculated per epoch and disbursed with proofs:

infra rewards claim --model summarizer-v1

Owners can integrate Webhooks or on-chain events for billing and analytics.

Summary

The InfraMind Model Marketplace is not just a directory—it’s a decentralized economy of deployable intelligence. Developers publish, the mesh runs, users invoke, and $IFM circulates. Whether you're building open-source tools, monetizing private agents, or responding to protocol bounties, the marketplace is your launchpad. Every model is portable. Every call is verifiable. Every listing is a proof that someone, somewhere, deployed code into the compute layer of a decentralized future.

PreviousGovernance & DAO ModelNextLinks

Last updated