MineAPI Whitepaper
The network design and official economic model for community-owned inference and the MMGA incentive layer.
Status: Whitepaper v0.2. The Tokenomics parameters in this document are the current official economic design. Verification implementation and the production BSC contract address will be published separately.
Abstract
MineAPI proposes a community inference network that turns fragmented idle compute into a useful API service. The system is designed around a simple belief: AI inference should not become an oligopoly game controlled only by a small number of infrastructure owners.
The network coordinates devices, routes model jobs, measures useful work and maintains two accounting streams: inference Credit and MakeMiningGreatAgain (MMGA) participation rewards.
The problem
Accelerators are frequently provisioned for peak demand but remain partly idle between experiments, training runs, creative work and local applications. At the same time, API buyers can face limited provider choice and opaque pricing.
Idle capacity alone is not a product. A usable network also needs:
- predictable model interfaces;
- device capability discovery;
- workload isolation;
- output and performance verification;
- reputation and availability records;
- transparent settlement.
MineAPI focuses on the coordination layer that turns heterogeneous devices into a consistent inference surface.
Video-first workload fit
MineAPI is designed to be professional in video models. Video generation is compute-intensive, parallelizable and frequently requested in bursts. That makes it a strong candidate for matching across capable consumer GPUs, creator workstations and Apple Silicon devices, rather than reserving GB300-class datacenter capacity for every generation.
The goal is not to treat all hardware as equivalent. The routing layer must enforce the memory, runtime, precision and performance requirements of each video tier while using community hardware where it is technically and economically appropriate.
Network design
Demand layer
API buyers use one account, one Credit balance and model-specific endpoints. They do not select a device owner directly. Requests declare the model, quality tier and generation parameters.
Routing layer
The router matches a request against memory, runtime, model cache, location, latency and reputation constraints. Routing should be model-aware and capable of retrying work when a host disappears.
Supply layer
Hosts run the MineAPI client on supported inference devices. A node advertises only currently available capacity and can stop accepting work whenever the owner needs the machine.
| Node state | Meaning | Reward eligibility |
|---|---|---|
| Inference | Processing an assigned request | Inference Credit plus eligible MMGA |
| Ready | Healthy and available for matching | Eligible MMGA uptime reward |
| Paused | Disabled by the owner | No active rewards |
| Offline | Health signal expired | No active rewards |
Proof and quality
A production network must not equate “job accepted” with “job completed.” The proposed design combines health attestation, deterministic checks where possible, output sampling, redundant execution for selected jobs, latency scoring and account reputation.
Verification requirements will differ by model family. Text token streams, diffusion/video outputs and deterministic preprocessing tasks should not share one simplistic proof rule.
Verification and privacy mechanisms remain active research and product-design areas. This frontend demo does not claim that these systems are already deployed.
MMGA on BSC
MakeMiningGreatAgain (MMGA) is the network incentive asset and is designed as a BEP-20 token on BNB Smart Chain. Its role is to bootstrap GPU supply and recognize eligible availability and participation even when short-term API demand is uneven.
MMGA does not replace inference income. Credit represents paid useful work; MMGA represents network participation under defined eligibility rules. Users connect a BSC-compatible wallet to receive on-chain rewards.
The Tokenomics below define its supply, emission and incentive structure. Before on-chain activation, MineAPI will publish the verified contract address and relevant contract information. Any MMGA market price displayed in the current frontend remains simulated preview data until a live market source is connected.
Tokenomics
Building a sustainable AI GPU network
MMGA is the core incentive asset of MineAPI’s decentralized GPU network.
Its operational utilities—including post-launch routing weight, reliability incentives, Burn-to-List model publishing, MMGA-denominated model transactions and governance—are defined in the MMGA Token Usage document.
Its purpose is not to replace real inference revenue. Instead, MMGA helps bootstrap GPU supply, maintain available compute capacity during periods of low utilization, and attract high-performance GPUs to the network.
As real inference demand grows, GPU Provider revenue should gradually shift from MMGA incentives toward real inference revenue.
Token incentives bootstrap the network. Real usage sustains it.
Key parameters
| Parameter | Value |
|---|---|
| Maximum Supply | 1,000,000,000 MMGA |
| Genesis Premine | 5% |
| Maximum Annual Emission | ≤ 10% / year |
| Target GPU Utilization | 20% |
| Early Staking Target APY | ~20% |
| Inference Revenue to GPU Providers | 85% |
Token supply
MMGA enters the ecosystem through two primary mechanisms.
Genesis premine
At network launch, 5% of Maximum Supply will be created as the Genesis allocation.
This equals 50,000,000 MMGA.
The Genesis Treasury may be used for:
- Early Staking
- Initial Liquidity
- Airdrops
- Community Events
- Ecosystem Launch
- Protocol Treasury
- Early Network Development
Ongoing emission
After launch, new MMGA issuance is capped at 10% of Maximum Supply per year.
For a 1 billion maximum supply reference, this means up to 100,000,000 MMGA per year.
This is a maximum emission limit, not a guaranteed inflation rate.
If GPU incentives, liquidity incentives, strategic allocations, or other ecosystem programs do not require their full annual budgets, unused MMGA will not be issued simply to reach the 10% cap.
Annual emission allocation
The maximum annual emission budget is allocated as follows:
| Category | Max Supply / Year | Share of Annual Maximum |
|---|---|---|
| GPU Network Incentives | 7.4% | 74% |
| Staking Rewards | 0.7% | 7% |
| LP Incentives | 0.4% | 4% |
| Strategic Investor OTC | 0.5% | 5% |
| Team & Protocol Treasury | 1.0% | 10% |
| Maximum Annual Emission | 10% | 100% |
The majority of new MMGA issuance is therefore directed toward building and maintaining the GPU network.
GPU network incentives
74% of annual emission is reserved for GPU incentives
GPU Providers are the core infrastructure participants of the network. They can earn through two separate revenue streams.
Real inference revenue
When a GPU completes real user inference workloads, the resulting revenue is distributed as:
- 85% to the GPU Provider
- 15% to the Platform
MMGA incentives
In addition to inference revenue, online GPUs may receive MMGA incentives.
These incentives primarily compensate for:
- Low-utilization periods
- Maintaining available GPU capacity
- Network cold start
- Idle capacity availability
- Early GPU network expansion
GPU utilization determines emission
GPU incentives are not released at a fixed rate regardless of network conditions. Instead, GPU MMGA emission adjusts according to the network’s effective GPU utilization.
Low utilization
Higher MMGA incentives help keep GPU capacity online.
Around 20% utilization
The network reaches its standard emission range.
Higher utilization
MMGA emission gradually decreases.
The decline remains intentionally moderate because higher utilization already creates several natural market effects:
- GPU Providers earn more real inference revenue.
- Higher revenue attracts additional GPUs.
- New GPU supply naturally dilutes MMGA rewards per GPU.
The protocol therefore does not need aggressive emission cuts.
| GPU Network Utilization | GPU Emission Multiplier |
|---|---|
| 0% | 1.50× |
| 5% | 1.375× |
| 10% | 1.25× |
| 20% | 1.00× |
| 40% | 0.95× |
| 60% | 0.90× |
| 80% | 0.85× |
| 100% | 0.80× |
20% represents the network’s target capacity utilization, not the point where MMGA incentives disappear.
Higher-performance GPUs earn more
GPU rewards are not calculated simply by counting the number of connected machines.
Each GPU receives a standardized performance score based on verified inference performance using a defined AI workload.
A reference GPU is assigned Reference Performance = 1.0.
| Verified Performance | Performance Score |
|---|---|
| 50% of Reference | 0.5 |
| Reference Performance | 1.0 |
| 150% of Reference | 1.5 |
| 200% of Reference | 2.0 |
GPU MMGA rewards are primarily determined by:
Verified Inference Performance × Effective Online Time
The network therefore rewards useful, available AI inference capacity rather than simply rewarding machines for being powered on.
Token price does not control emission
MMGA market price is not used as an input to the emission mechanism.
The protocol does not increase issuance when MMGA price falls. It also does not reduce issuance simply because MMGA price rises.
There is no:
- Price Compensation
- Token Price Peg
- Automatic Emission Increase During Price Declines
- Automatic Emission Reduction During Price Rallies
GPU Providers receive MMGA according to network conditions and verified performance. The market determines the economic value of those MMGA rewards.
Staking
MMGA holders can participate in the network through Staking.
Staking is designed to support:
- Governance participation
- Long-term ecosystem alignment
- Lower short-term circulating supply
- Additional MMGA rewards
During the early network phase, the Team is expected to represent a significant portion of total Staked MMGA.
Early target APY
The early Staking target is approximately 20% MMGA APY.
This is not a permanently guaranteed yield.
Maximum annual Staking emission is capped at 7,000,000 MMGA per year.
As more MMGA enters the Staking system, APY naturally decreases.
| Total Staked | Estimated MMGA APY |
|---|---|
| 20M | 20% |
| 35M | 20% |
| 50M | 14% |
| 70M | 10% |
| 100M | 7% |
This allows Staking rewards to adjust naturally as participation grows without requiring constant manual interest-rate changes.
Liquidity
MMGA will maintain public market liquidity.
The Genesis Treasury may provide part of the initial MMGA liquidity required for launch.
In addition, up to 4,000,000 MMGA per year may be allocated to external Liquidity Provider incentives.
LP incentives are not guaranteed emissions. If market liquidity and trading-fee revenue are already sufficient, actual LP incentives may be reduced. Unused LP allocation does not need to be issued.
Strategic investor OTC
Up to 5,000,000 MMGA per year may be allocated to strategic OTC transactions.
Potential participants may include:
- Strategic Investors
- GPU Infrastructure Partners
- AI Model Partners
- Ecosystem Partners
- Long-Term Strategic Capital
Strategic OTC MMGA must enter a Vesting Contract.
A baseline structure may include a 6-Month Cliff followed by 18-Month Linear Vesting.
Locked OTC MMGA is not considered part of the Liquid Float. This allows strategic financing without creating immediate market sell pressure.
Team & protocol treasury
In addition to the Genesis Premine, up to 10,000,000 MMGA per year may be allocated to the Team and Protocol Treasury.
These MMGA may support:
- Protocol Development
- Team Incentives
- Infrastructure
- Ecosystem Development
- Governance Treasury
- Long-Term Network Growth
Team and Treasury emission is intended to follow actual ecosystem growth.
If the network does not use the full annual emission budget, the Team does not need to issue its full annual allocation independently.
Launch phase
The first 180 days after network launch are considered the Launch Phase.
During this period, the protocol may require rapid adjustments to:
- GPU Benchmarks
- Oracle Infrastructure
- GPU Reward Parameters
- Staking Parameters
- Emission Parameters
- Security Mechanisms
For this reason, the Team and Protocol Treasury are designed to maintain strong governance and MMGA control during the initial phase.
The target is Team / Treasury Controlled MMGA ≥ 50% of Circulating Supply during the first 180 days.
This Launch Phase constraint expires after 180 days.
Over time, as GPU Providers earn MMGA, community Staking expands, liquidity increases and strategic allocations unlock, the Team’s share of circulating supply and governance power is expected to decline naturally.
Circulating supply is not the same as total supply
To provide transparent supply information, the ecosystem distinguishes between three concepts.
Total supply
All MMGA that have been issued.
Circulating supply
MMGA that have been unlocked and are available for use or transfer.
Liquid float
MMGA that are realistically available for unrestricted market trading.
Therefore:
- Strategic Investor Vesting MMGA are not part of Liquid Float.
- Unused Treasury MMGA are not part of Liquid Float.
- Staked MMGA are not part of Liquid Float.
This distinction provides a clearer picture of actual market supply.
The economic loop
Early network stage
Low GPU utilization
→ Higher MMGA incentives
→ More GPUs remain online
→ The network maintains sufficient inference capacity
Growing real demand
GPU utilization increases
→ GPU Providers earn more real inference revenue
→ More GPUs join the network
→ MMGA rewards are naturally diluted across a larger GPU supply
→ Dependence on MMGA subsidies gradually decreases
Long-term objective
The objective is not to make GPU Providers permanently dependent on MMGA rewards.
A sustainable GPU network should ultimately be driven by:
Real Inference Demand
→ Real GPU Revenue
→ More Efficient GPU Supply
→ Better Pricing and Better Inference Experience
MMGA primarily helps the network cross the difficult cold-start stage and maintain sufficient available compute capacity as the ecosystem develops.
Token incentives bootstrap the network. Inference revenue sustains the network. Real usage creates long-term value.
Roadmap
- Frontend preview — product language, account flows, calculators and node dashboard.
- Closed host test — client telemetry, capability discovery and controlled model jobs.
- Verified inference pilot — settlement ledger, failure handling and reputation signals.
- Community network beta — broader device support, public API access and transparent metrics.
- BSC reward activation — audited contract, published token parameters and wallet withdrawals.