$LLAMA Tokenomics

A standard Sui coin created with coin::create_currency and a one-time witness: 100,000,000 $LLAMA, 9 decimals — matching the MIST convention so every amount stays integer math end to end.

Players earn the right to mint by playing, so the TreasuryCap cannot simply be burned at genesis the way a fixed-supply airdrop token's would be. Instead it is sealed inside a shared mint controller that enforces the ceiling on-chain: every mint checks minted + amount ≤ max_supply, and max_supply can only ever be lowered, never raised.

The practical guarantee is the one a burned cap gives — nobody, the house included, can push supply past the published number — while still letting minting happen over time as players earn it. Be precise about what that does and does not cover: the ceiling is trustless, the pace underneath it is not. Minting needs a separate capability from the admin one, and the copy of it that exists today lives inside the claim contract, which will only release tokens against a published Merkle root. But the admin can create another. What that means in practice is set out in §03 rather than left for you to find in the source.

The package was published to Sui mainnet on 29 July 2026 and is immutable from that moment — a Sui package cannot be edited, only superseded by a new one at a new address. The 100,000,000 ceiling and the halving multipliers [16, 8, 4, 2, 1] were fixed by that transaction and cannot now be changed by anyone, us included.

The same day, 57,000,000 $LLAMA — 57% of supply — was minted directly into five irrevocable vesting locks. It never passed through a spending wallet: the mint and the lock happen in one transaction, so there was no moment at which anyone held it loose. The locks are listed with their live released figures on the contracts page.

Nothing below this line has been switched on. The emission schedule has issued 0 of its 25,000,000 play-to-earn pool, the vault holds 0 staked $LLAMA, and the liquidity pool does not exist yet. Seasons, rates and the airdrop described in the rest of this page are the design the deployed contracts implement — they are not yet things you can go and do.

Everything above describes what is now beyond our reach. This section is the other half, and it is the half most token pages leave out: the powers that still exist, who holds them, and what would happen if that holder were compromised or dishonest. None of it is hidden — all of it is readable in the Move source and on the explorer — but you should not have to go looking.

One key, not a multisig — yet

The AdminCap was moved off the production server on 29 July 2026 to a cold address whose key is not on any machine that runs the casino. Control of that address was proven by a round trip before the capability was sent to it, and the transfer script refuses any destination derived from a key in the server's own keystore — otherwise the move would be theatre.

It is still a single key. A multisig is the right home for it and is the next step, but until that happens the honest description is one signer, held offline, with the losses above possible if it is compromised — and everything above it permanently forfeited if it is lost. You can watch the address: 0xee63ebef00…33915e48 ↗

The vesting beneficiaries are single keys, permanently

Each lock pays a fixed address that was set when the lock was created and can never be changed — there is no setter, by design, because a changeable beneficiary is not really a lock. The same immutability that stops us moving tokens early also means a beneficiary key that is lost takes its allocation with it, and one that is compromised hands the attacker that stream as it releases. Five separate keys hold the five locks; none of them is a multisig today.

Two unlock cliffs arrive in one step

Cliffs are not gentle. A 12-month cliff means nothing releases for 12 months and then the whole first year vests at once — so 3,750,000 $LLAMA becomes releasable to the team in a single moment on 23 July 2027, and 4,000,000 to the bankroll reserve on 17 July 2028. After each step the remainder resumes releasing continuously. The contract counts a month as 30 days, which is why those dates are not anniversaries of the start date.

The three locks with no cliff — treasury, vault bootstrap and ecosystem — have been releasing since day one, at roughly 28,000 $LLAMA a day between them. That is the only supply that can move at all right now, and it is house-side: no player can claim anything until a Merkle root is published.

Burning frees headroom rather than lowering the ceiling

The sinks in §10 burn $LLAMA, and burning decrements minted — so it genuinely reduces circulating supply, but it also returns that room under the cap where a minter could reach it again. The alternative would make sinks permanently deflationary at the cost of shrinking the programme every time someone bought compute credit. lower_max_supply is the one-directional tool for actually retiring supply, and it is the honest thing to call once the distribution is finished.

What would make this section shorter

Three things, in the order they are worth doing: move the AdminCap to a multisig; call revoke_minter once emissions, liquidity and the airdrop are funded, which permanently ends discretionary minting; and call lower_max_supply down to whatever has actually been distributed. Each one is a one-way transaction that removes a power from us, and each one will be announced with its digest so you can verify it happened rather than take our word for it. None of them has been done yet.

Weight sits on future play rather than past play. The retroactive airdrop is deliberately a small slice: measured lifetime real-money volume on Llamabet so far is modest, and a large retro allocation would set a price anchor the platform cannot grow into while handing a big share of supply to a handful of wallets.

You earn on what you actually contributed to the house, not on how much volume you could churn. Every game has a different edge, so the same stake generates very different revenue — and paying on raw volume would make grinding the lowest-edge game the optimal farm. Llamabet already records revenue contributed on every settled bet:

Poker is measured rather than estimated. Because the house takes rake instead of an edge, the poker server records your actual share of the rake in every pot you contribute to, split in proportion to what you put in — the standard “contributed” method used for rakeback.

Note what is not in that formula: anyone else. The rate is fixed, so your allocation depends only on your own play. An earlier draft split a fixed pool pro-rata, which meant your rate quietly fell as other people played — impossible to publish in advance, and it rewarded hoping the casino stayed empty.

The rate halves as each fifth of the play-to-earn pool is issued, so the earliest players earn 16× what the last ones will. It decays with tokens issued, not elapsed time — a quiet month cannot burn through the opening rate, and no amount of waiting can game it.

The rate is linear in revenue contributed, on purpose. Softening the curve to favour smaller players sounds fairer but is trivially gamed: sign-in derives a wallet with no seed phrase, so splitting across ten wallets would simply pay more. Linear splitting changes nothing, which is what makes it sybil-neutral. A per-wallet cap and a minimum-activity floor limit concentration instead.

Referrals

Refer a player and you earn a 20% bonus on everything they earn, for as long as they play — at the opening rate, 3.2 $LLAMA for every SUI of revenue they generate.

It is a bonus, not a cut. Being referred never reduces what the referee earns; they keep their full rate. The programme exists to bring players in, so taking rewards from one player to pay another would defeat the point.

That does mean referred play emits 1.2× the rate — and since sign-in derives a wallet with no seed phrase, anyone can refer themselves from a second wallet. Rather than police that, the cost is carried by the emission budget: it is set at 20% of house revenue precisely so the bonus is affordable while self-referral still loses money. See the emission rule.

A separate, one-off 3% of supply for players who were here before the token existed. It is not the same mechanism as the rate above, and the difference is worth being explicit about.

Derived from the rate, a retroactive airdrop would be worth almost nothing: all pre-launch play across every player and every game amounts to roughly 13 SUI of house revenue, which at the opening rate is about 208 $LLAMA in total. So the airdrop is a fixed grant split pro-rata among pre-launch players instead — the one place pro-rata is the right mechanism, because it is a gift rather than a published rate, and nothing was promised in advance for it to float against.

Claims work by publishing a Merkle root on chain and letting each eligible wallet claim against it, so the recipient list costs one on-chain write rather than thousands. Roots are write-once per epoch, so an allocation cannot be revised after players have started claiming against it. Unclaimed tokens return to the treasury when the window closes.

The eligibility snapshot is taken at a block height in the past and published with the announcement. A forward-looking snapshot would simply invite wash trading into it. No snapshot date is published yet, and none will be until it has already passed.

Seasons, not one drop

The retroactive grant is Season 0 of several. One snapshot rewards whoever happened to be there and leaves everyone who arrives afterwards with nothing to play for; seasons keep a live reason to play, and step the rate down each time so being early is worth something real rather than merely promised.

Rates are per SUI of house revenue you generate, not per SUI wagered. Seasons advance on tokens issued, never on a date, so a quiet month cannot burn one down. Snapshot and start dates are not published yet.

A revenue-share token that can only be farmed shuts out everyone who arrives after the farm slows down; one that can only be bought shuts out the players who generated the revenue in the first place. Both routes exist, and the vault treats them identically — a staked token earns the same dividend whether it was earned at a table or bought on a DEX.

The buy route is what the 15% protocol-owned liquidity allocation is for. It is paired into a Sui DEX pool at launch and its LP tokens are locked, so the market has depth from day one and that depth belongs to the protocol rather than to a market maker who can withdraw it. Anyone can buy there and stake straight into the vault without ever having placed a bet.

Stake $LLAMA and receive a pro-rata claim on the house-profit distribution, claimable from your own wallet whenever you choose. The important detail: the dividend is paid in SUI, not in $LLAMA. It is a share of real revenue rather than emissions wearing a costume.

Once a day the house funds the vault in a single transaction. It does not compute a list of stakers, calculate anybody's share, or send per-player transfers — the vault does all of that on-chain. Your entitlement for the day is simply:

Active is the operative word. Stake still warming up, and stake being unbonded, count for neither — they earn nothing and dilute nobody. Entitlement then accumulates in the vault until you withdraw it, so being away for a month costs you nothing and there is no deadline to claim by. Internally it is an accumulator, so claiming is constant-time however many stakers exist and the house pays gas once per day rather than once per player.

Two safeguards are already live rather than planned. Distributions accrue against a high-water mark, so a losing day must be earned back before new dividends flow; and payouts are released from a pool a fraction at a time, so a single lucky day cannot drain the bankroll and a quiet stretch does not stop rewards dead. Without both, a share of “up days only” pays out a share of variance rather than a share of profit — on a bankroll that goes 1,000 → 900 → 1,000 the house has earned nothing, yet a naive daily rule still pays out on the recovery.

One rule decides whether a casino token survives. If the tokens earned by wagering are worth more than the house edge, then wagering becomes a profitable farm, volume turns fake, and the house funds its own extraction. So emissions are budgeted in value terms, at 20% of the revenue that generated them. That budget holds exactly averaged across the whole schedule — but it is front-loaded, and Season 1 spends more than its share. The 16× opening rate is a deliberate decision to pay early players above the rule and claw it back later:

Season 1 is the shortest season for exactly this reason: it ends once the first fifth of the pool is issued, which is about 3% of the programme's lifetime revenue. A per-wallet cap limits what any single address can take out of one epoch, and every season after the first sits below the rule — halving each time — so the average comes back to 20%. The trade is a real one and it is stated here rather than buried: the earliest players are paid above cost, and that cost is borne by the schedule, not by later players' rates, which are fixed in advance.

In practice this runs as a SUI-denominated daily budget converted at a time-weighted average price, so a rising token price emits fewer tokens and extends the runway automatically — paired with a hard per-epoch token ceiling so a price crash cannot drain the pool.

Emissions without demand that consumes tokens is just a slow sell. These sinks are things Llamabet already charges for, so they burn tokens against real costs rather than invented ones:

These are the numbers that decide whether the design is working. They are published so the commitment is checkable rather than asserted.

The dividend scales with the casino, and the casino is early. A share of house profit is only worth what the house earns. At current volume that figure is small, and any yield expectation should be set against real revenue rather than a projection. The token is a distribution and loyalty instrument first; the profit claim becomes meaningful only if volume grows substantially.

Regulatory. A token carrying a share of business profit may be treated as a security in some jurisdictions. Access to the claim and the vault will be geo-gated at least as strictly as gameplay already is, and the design may shift toward buyback-and-burn instead of a direct dividend if that becomes the sounder footing. This page is not an offer, and it is not financial or legal advice.

Key custody. The AdminCap is a single offline key rather than a multisig, and each vesting lock pays a single beneficiary key that can never be changed. Losing one forfeits what it controls permanently; compromising one costs what §03 sets out. This is the largest remaining trust assumption in the design and it is being worked on rather than solved.

Not everything is deployed. Supply, vesting and the vault are live and immutable. The emission seasons, the retroactive airdrop and the protocol-owned liquidity are not switched on yet — rates and dates for those may still change, and none of them is a promise you can hold anyone to today.

No audit yet. The contracts have unit tests and were rehearsed end to end on testnet before the mainnet publish, but no third-party security audit has been completed. That is stated here rather than implied by silence.