Your cost curve flattens
Metered APIs charge for every thousand tokens, so the better your product gets, the more you owe. A staked allowance is fixed: the ten-thousandth prompt of the day costs exactly what the first one did.
$0 marginal cost
Stake $GROWE and mint a permanent daily allowance of zero-retention inference across text, image, audio and agents. The metered bill goes to zero. The position stays yours.
You hold
72,000 $GROWE
staked, never spent
Every 24h
$1.00 of inference
refills automatically, forever
$GROWE · the mechanism
Most AI tokens bolt a currency onto a product. Growe inverts it: the token is the subscription. Holding is the payment method, and the payment is refundable.
Stake
72,000 $GROWE
Locked, not consumed
Mint
1 credit / day
≈ $1.00 of inference
Unlock
72,000 $GROWE
Same balance, after the cooldown
Net cost of a year of AI = the opportunity cost of the stake, not the stake itself
Metered APIs charge for every thousand tokens, so the better your product gets, the more you owe. A staked allowance is fixed: the ten-thousandth prompt of the day costs exactly what the first one did.
$0 marginal cost
Paying an invoice destroys capital. Staking parks it. The stake keeps minting inference for as long as it stays locked, and it is still there when you unlock.
100% capital retained
Credits regenerate every 24 hours at the live mint rate. There is no top-up, no overage invoice and no billing surprise at the end of the month.
Refills every 24h
−0.0417%
Mint rate decay, per epoch
Supply is fixed at one billion and cannot be increased, so the allowance is not printed — it is funded by trading fees that buy GPU capacity. As capacity grows, the $GROWE required to mint one daily credit falls on a published schedule. Stakers who are early keep the cheaper rate for the life of their position, and the same allowance costs progressively more of a supply that never grows.
How it works
One Uniswap V3 pool on Robinhood Chain, paired with WETH, liquidity locked at deploy. No presale to miss, no allowlist, no migration to wait for.
Locked $GROWE mints daily inference credits at the live mint rate. The stake is never spent and never burned — it keeps producing while it stays locked.
Credits refill every 24h across the app and the API. One key, every modality. Unused credits expire; the allowance does not.
Run the numbers
Move the slider to your current monthly spend on AI inference. Everything below is derived from the live mint rate and spot price.
Five-year comparison
Metered API
$30,000
Spent and gone. Grows with every request.
$GROWE staked
$5,141
Deployed once. Still held at the end of year five.
Breakeven
10.3 mo
then inference is free
Avoided over 5 yrs
$30,000
minus the retained stake
Illustrative only. Assumes a $0.0042 spot price, a 72,000 $GROWE mint rate and a constant workload. Token price moves both ways — the allowance does not.
Get $GROWE on PONSSide by side
Same models, same throughput. The difference is what you own at the end of the year.
Pricing model
Per 1K tokens, per request
Fixed daily allowance
Cost trajectory
Scales up with every prompt
Flat — usage growth is free
Capital at the end
Zero. Spend is gone
Stake retained, still minting
Rate limits
Tier-gated, opaque
Set by your own stake
Data retention
Logged, analyzed, trained on
Zero retention, no logs
Provider lock-in
One vendor, one API
Every open model, one key
$GROWE is a utility asset for accessing compute. It is not a security, a yield product, or a claim on revenue.
Launch · PONS
$GROWE deploys through the PONS factory on Robinhood Chain: the entire supply is minted once and paired into a single locked Uniswap V3 pool in the same transaction. There is no treasury bucket to unlock and no mint function to abuse.
The PONS pool charges a 1% fee on each trade. It accrues inside the locked liquidity position — nobody can withdraw the liquidity itself.
1% pool fee
PONS routes 70% of that fee to the token creator and 30% to the protocol. Our share goes straight into inference capacity — the treasury buys compute, not marketing.
70% creator share
Capacity bought with fee revenue sets how many credits the network can mint. More volume means more GPUs, which means the mint rate can fall and every staked token buys more inference.
Volume → capacity
Launch parameters
Fair launch window
PONS caps the opening blocks: only the creator's initial buy clears in the launch block, and in the next one no wallet may hold more than 5% of supply or buy more than 5.5%. Selling stays open the whole time. After the window, the limits are gone for good.
Liquidity is locked by the PONS locker at deploy, so the pool cannot be pulled. Fees accrue inside that locked position — 70% of them are what pays for the compute behind your allowance.
Settlement layer
An allowance that refills daily needs a chain where daily accounting is effectively free. Robinhood Chain settles in ETH with sub-cent fees and fast finality, and PONS gives us a launch where the liquidity is locked from the first block.
Credits accrue and settle every epoch. Anywhere with real gas, the fee would eat the allowance; here it rounds to nothing.
Stake, unstake and mint-rate updates confirm in seconds, so the allowance you see is the allowance you have.
The chain's native asset is ETH and the pool is paired with WETH. No bridge token, no synthetic quote asset to price.
Chain id 4663, Uniswap V3 pools, an EVM RPC. Any wallet, indexer or bot that speaks Ethereum already speaks this.
One key, every modality
Chat, images, speech and tool-calling agents all draw from the same daily allowance. OpenAI-compatible, so existing code works with a base URL swap.
Reasoning, long-context chat and code with open-weight frontier models.
Generation, editing and upscaling. Uncensored, watermark-free output.
Speech synthesis and transcription in 40+ languages.
Function calling, web search and structured output for autonomous stacks.
from openai import OpenAI
client = OpenAI(
base_url="https://api.growe.green/v1",
api_key="your-growe-api-key",
)
resp = client.chat.completions.create(
model="growe-large",
messages=[{"role": "user", "content": "Summarize this contract"}],
)Privacy architecture
Prompts are not logged, not retained and not used for training. Pick how far up the stack that guarantee is enforced in hardware.
Tier 1
Requests to third-party models are stripped of every identifying header before they leave our edge.
Tier 2
Open-weight models run on our own GPUs. Prompts are held in memory for the length of the request, then dropped.
Tier 3
Hardware enclaves execute inference in a sealed environment. Growe itself cannot read your session.
Tier 4
Client-side encryption. Your prompt is ciphertext until it reaches the enclave that decrypts it.
Questions
$GROWE launches through PONS on Robinhood Chain — fixed supply, locked liquidity, no allocation. Follow the account for the contract address and the launch block.