Ctrlk
For the complete documentation index, see llms.txt. This page is also available as Markdown.

💿Bittensor SRB

Learn how Bittensor staking rewards are calculated. TAOSRB formula covers root subnet delegation, observed root APY, validator take, and why alpha staking is measured separately.

TAOSRB is a benchmark representing the mean, annualized staking rate earned on TAO delegated to Bittensor validators on the root subnet (Subnet 0).

TAOSRB is calculated and published by Staking Rewards via the Bittensor Profile and Data API.

TL;DR: On Bittensor, staking TAO means delegating it on the root subnet. TAOSRB is the root-stake-weighted mean of each validator's observed root APY, net of the validator's take. Alpha staking — swapping TAO for a subnet's alpha token — is a different product with returns denominated in alpha, and is deliberately excluded from TAOSRB.

TAOSRB=∑v=1napyv⋅rsv∑v=1nrsvTAOSRB = \frac{\sum_{v=1}^{n} apy_v \cdot rs_v}{\sum_{v=1}^{n} rs_v}
Formula key
Metric name
Description

apy_v

Root APY (v)

The observed 7-day annualized return earned by TAO delegated to validator v on the root subnet, net of validator v's take.

rs_v

Root Stake (v)

TAO delegated to validator v on the root subnet (Subnet 0).

n

Active Root Validators

The number of validators registered and earning on the root subnet.

  • TAOSRB is non-compounded and excludes returns from reinvestment of rewards.

  • TAOSRB includes both claimed and unclaimed rewards.

  • TAOSRB is net of validator take, because the observed root dividend a delegator receives is already after the validator's commission.

  • Only validators with a measurable root return are included. Operators commonly keep secondary or dust hotkeys registered on root that never enter an epoch and earn nothing; these are excluded rather than published as 0% options. The included set holds over 99.9% of all root stake.

Why alpha staking is excluded

Bittensor offers two distinct staking products, and blending them produces a number that describes neither.

Root stake and alpha stake do not earn at the same rate. Within each subnet's emission weighting, root stake is discounted by a TAO weight factor of 0.18 while alpha stake counts at 1.0. Root APY therefore runs at roughly 0.18x the corresponding subnet APY — currently ~4-7% against subnet yields of ~30-48%.

A combined ratio of the form

total nominator returnrsv+asv\frac{\text{total nominator return}}{rs_v + as_v}

is consequently governed by how much of a validator's book its delegators happen to hold in alpha, not by the validator's performance. Two validators offering an identical root APY will show wildly different figures purely because of their stake composition, and neither figure is what a new TAO delegator would earn.

The two products are also denominated differently. A root dividend originates as alpha on the subnet that generated it, but is auto-sold to TAO at accrual, so a root return is denominated in TAO and settles to TAO on redemption. An alpha staking position is not: the staker holds the subnet's alpha token, and converting back to TAO means selling into that subnet's AMM pool, where size moves the price. Marking an alpha position's emission to the spot rate and reporting it as a TAO yield therefore overstates what is realisable.

This is worth stating precisely, because it is easy to misread. "Alpha" names two different things here:

  1. The form root rewards are generated in. A root staker's rewards are emitted as alpha across every subnet their validator is active on, and the chain sells that alpha to TAO each block. TAOSRB includes these — they are the root reward.

  2. Alpha staking, a separate product. TAO is swapped through a subnet's AMM into its alpha token, and the alpha is staked. Returns are paid and held in alpha, which can fall against TAO. TAOSRB excludes this.

So TAOSRB does not exclude "rewards paid in alpha". It excludes the alpha staking product.

Subnet yields remain a real and material part of the Bittensor ecosystem. They are an alpha-denominated return on a different asset, and are not represented by TAOSRB.

Bittensor Network Architecture: Dynamic TAO (dTAO)

Since February 2025, Bittensor operates as a multi-subnet network under the Dynamic TAO (dTAO) model. Each subnet has its own alpha token with an AMM (Automated Market Maker) pool that facilitates exchange between TAO and the subnet's alpha token.

Staking modes:

  • Root Staking (Subnet 0): TAO is delegated directly to validators on the root subnet. Root itself holds no alpha token. A validator's root dividends are generated as alpha on each subnet it validates and are then auto-sold to TAO, so the return is denominated in TAO. Root stake contributes to a validator's weight with a TAO weight factor of 0.18. Under Root Reborn (live on mainnet since 4 August 2026, runtime spec_version 441+) dividends accrue into a per-validator escrowed basket and are held until the staker calls claim_root; since 28 August 2026 validators also curate which subnet alphas that basket keeps — see the note below.

  • Alpha Staking (Subnets 1+): TAO is exchanged for subnet-specific alpha tokens via the AMM pool. Returns are earned in alpha tokens, which carry additional price risk as the alpha/TAO exchange rate fluctuates.

A validator's total effective stake (in TAO terms) is: root_stake + alpha_stake_as_tao, where alpha_stake_as_tao is the TAO-equivalent value of all alpha tokens staked with the validator across subnets.

Root Reborn is live; validator curation of baskets is not yet. Root Reborn shipped to mainnet on 4 August 2026 at runtime spec_version 441, and the chain reports 448 at the time of writing. It rolled out in two stages, and only the first is currently active.

Live now — escrow and manual claim. Root dividends are no longer compounded onto root stake automatically. They accrue into a per-validator escrowed basket, and a staker's balance no longer rises on its own: rewards are realised only through a manual claim_root, which values basket shares at current NAV, redeems pro-rata across every validator the staker is with, and settles to TAO. Claiming costs a fee that scales with basket width, and dividends accrue roughly every two days and somewhat randomly. This stage is directly observable on chain — every root claim now carries a manual-claim extrinsic, where none did before the rollout.

Live since 28 August 2026 — basket curation. Each validator now publishes a weight vector determining which subnet alphas its basket keeps, with the rest sold to buy more of the kept ones. The on-chain gate SubtensorModule.RootWeightSettingEnabled flipped true at runtime 450. The basket is therefore a product validators compete on alongside their take, and root returns can begin to diverge on basket strategy — a source of genuine per-validator variation that did not exist before.

That divergence has not yet appeared in the published rates, because the seven-day observation window still consists mostly of pre-curation data. Measured across validators holding more than 5,000 TAO of root stake, the spread was 6.08 percentage points the day before curation and 5.18 the day after. The asset script tracks this spread every run and flags it when it widens materially, since that is the point at which the benchmark's treatment needs revisiting.

What this means for TAOSRB. The benchmark reports each validator's observed root return, so divergence driven by basket strategy is captured as it emerges rather than needing a formula change. Two caveats stand. The rate measures value at accrual: it does not capture movement in the escrowed basket between accrual and the staker's eventual claim_root, nor the claim fee, so a staker who claims promptly realises approximately the published rate while one who leaves rewards escrowed takes on that exposure in either direction. And now that baskets hold curated alpha positions, the extent to which a basket's value growth is a TAO yield rather than alpha price movement is an open question, pending confirmation of how the underlying data source values post-curation accrual.

Emission Model and Halving Schedule

Bittensor emits new TAO tokens with each block (~12 seconds per block, 7,200 blocks per day). The block reward follows a halving schedule tied to total supply issuance (not block height):

Halving
Trigger (Total Issued)
Block Reward
Daily Emission
Annual Emission

Pre-halving

< 10,500,000 TAO

1.0 TAO

~7,200 TAO

~2,628,000 TAO

1st (Dec 2025)

10,500,000 TAO

0.5 TAO

~3,600 TAO

~1,314,000 TAO

2nd (future)

15,750,000 TAO

0.25 TAO

~1,800 TAO

~657,000 TAO

Maximum supply is capped at 21,000,000 TAO. The emission rate is determined dynamically based on total issued supply.

Emission distribution within subnets follows a flow-based allocation model (since November 2025), where each subnet's share of total emissions is determined by net TAO staking inflows, smoothed with a 30-day half-life exponential moving average. Within each subnet, emissions are split: 41% to validators/stakers, 41% to miners, and 18% to subnet owners.

Slashing

TAOSRB does not factor in slashing. Bittensor does not implement protocol-level slashing for validators or delegators. Some individual subnets may implement miner collateral slashing mechanisms, but these are subnet-specific and do not affect validator or nominator stake.

TAO Validator Staking Reward Rate: TAOSRB(v)

Staking Rewards publishes each root validator's observed root APY directly.

TAOSRB(v)=apyvTAOSRB(v) = apy_v
Formula key
Metric name
Description

apy_v

Root APY (v)

The observed 7-day annualized return on TAO delegated to validator v on the root subnet, measured from realized root dividends and already net of validator v's take.

Because the figure is measured from realized dividends rather than derived from stake weights, no separate commission adjustment is applied: a validator's take is already reflected in what its delegators received. Differences between validators therefore reflect what actually differentiates them: their take, the consistency of their epoch participation, and — since basket curation went live on 28 August 2026 — the composition of the basket their root dividends are held in. In current conditions this places active root validators in a band of roughly 4% to 7%.

How the root APY is measured

apy_v is not a modelled figure. It is built from the per-block dividends a validator's root nominators actually received:

apyv=ndvrsv⋅tyapy_v = \frac{nd_v}{rs_v} \cdot t_y
Formula key
Metric name
Description

nd_v

Nominator Dividends (v)

Root dividends credited to validator v's nominators over one Bittensor day (7,200 blocks), averaged over the trailing 7 days. Accounts for hotkey swaps and parent/child hotkey arrangements.

rs_v

Root Stake (v)

TAO delegated to validator v on the root subnet over the same window.

t_y

Year Constant

365 days.

Root dividends are themselves a consequence of how root stake enters subnet emission weighting. Each subnet splits a validator's dividends between its root and alpha stakeholders by the root proportion:

rp=taoroot⋅wtaotaoroot⋅wtao+αissuedrp = \frac{tao_{root} \cdot w_{tao}}{tao_{root} \cdot w_{tao} + \alpha_{issued}}

where tao_root is the TAO staked on root, α_issued is all alpha the subnet has emitted (alpha_in + alpha_out), and w_tao is the chain's TAO weight parameter, currently 0.18. Root stake therefore earns approximately 0.18x the rate that alpha stake earns in the same subnet, and the root proportion declines over a subnet's life as its alpha supply grows.

That factor is what separates the two rates: root APY sits near 4–7% while subnet APYs sit near 30–48%. Deriving root APY from the stake weights alone —

apyv≈nrv⋅ty⋅wtaowsvapy_v \approx \frac{nr_v \cdot t_y \cdot w_{tao}}{ws_v}

where nr_v is the validator's daily nominator return and ws_v its total weighted stake — reproduces the observed root APY to a stake-weighted 6.20% against 5.55% measured, confirming the measured figure independently.

Observation Period:

  1. TAOSRB is determined from a trailing 7-day window of realized root dividends at the point of evaluation.

  2. That observation is annualized for a comprehensive year-long projection.

A 7-day window is used rather than a single day because root dividends are not awarded evenly. The chain awards them roughly every two days and somewhat randomly, and a week between accruals is not abnormal. A single-day reading is therefore dominated by whether an accrual happened to land in it. A 7-day window absorbs that without materially lagging changes in network conditions.

For the same reason, a validator with very little root stake can have one accrual dominate even its 7-day window. Validators whose 1-day and 7-day rates diverge by more than a factor of two are treated as an unstable observation and excluded until they settle; in practice every validator with meaningful root stake sits within a few percent of 1.0.

Calculation

The calculation happens every 2 hours and is immediately published via Data API.

Limitations

  1. TAOSRB employs a 365-day convention for annualizing returns. This convention remains consistent and does not accommodate adjustments for leap years.

  2. TAOSRB covers root staking only. It is not an estimate of the return on alpha staking, which is denominated in subnet alpha tokens and carries alpha/TAO exchange rate risk.

  3. TAOSRB is a root-stake-weighted network average. It is dominated by the largest root validators and is not a guarantee of any individual validator's rate.

  4. Self-staked tokens are not currently distinguishable from delegated tokens in the validator data source.

  5. Root return depends on the flow-based allocation of emissions across subnets, which shifts with staking activity, so TAOSRB moves with network conditions and not only with validator behaviour.

  6. Validators registered on root without a measurable return are excluded rather than reported at 0%. They hold under 0.1% of root stake.

TAO Real Reward Rate: TAOSRB^R

The real reward rate calculates the TAOSRB adjusted for inflation in the network.

TAOSRBR=1+TAOSRB1+ir−1TAOSRB^R = \frac{1 + TAOSRB}{1 + ir} - 1
Formula key
Metric name
Description

ir

Inflation Rate

The annual rate at which TAO supply increases, calculated as annual emissions divided by circulating supply, which is TAO minted to date as read from the chain.

Inflation Rate Behavior

TAO inflation is driven purely by block rewards. Both sides of the ratio are taken from the chain's own issuance figure: the halving tier that sets the block reward is read off total issuance, so the denominator is the same quantity rather than a third-party estimate. The inflation rate decreases over time due to:

  1. Halvings: Block rewards are cut in half at predetermined supply thresholds (first halving at 10.5M TAO issued in December 2025).

  2. Growing supply: As the denominator grows, the same emission rate produces a lower inflation percentage.

  3. Supply cap: Maximum supply is 21,000,000 TAO. Recycling of TAO from subnet registration fees returns it to the unissued pool, which both slows net issuance and delays halving events.

Other Bittensor metrics calculated by Staking Rewards:

Metric
Description

Delegated Tokens

The total TAO tokens delegated to validators, including both root stake and alpha stake (in TAO equivalent). Note this covers both staking products, while TAOSRB covers root only.

Self Staked Tokens

The number of TAO tokens staked by validators directly. Currently not distinguishable from delegated tokens in the data source.

Staking Wallets

Total number of unique delegations on-chain across all validators. It doesn't distinguish based on unique delegator addresses.

Staked Root

TAO staked directly on the root subnet (Subnet 0).

Staked Alpha

TAO staked via alpha tokens across all subnets (in TAO equivalent based on AMM exchange rates).

Total Supply

The 21,000,000 maximum supply.

Circulating Supply

TAO minted to date, read from the chain rather than from a third-party estimate. Currently about 53% of maximum supply.

Staking Ratio

Staked tokens over circulating supply. Staked tokens covers both staking products, counting alpha stake at its TAO-equivalent value.

Reward Frequency

Two days, the cadence at which root dividends accrue into the validator's escrowed basket. The chain awards them irregularly and a week between accruals is not unusual. Note this is accrual, not receipt: dividends reach the staker's balance only on a manual claim_root.

Lockup Time / Unstaking Time

Both zero. Root delegation carries no committed term, and unstaking settles immediately. Bittensor's conviction locks bind alpha stake in subnet-ownership contests, not root delegation, so they do not apply to the reward options covered by TAOSRB.

Data Sources:

  • Bittensor root subnet returns and network data via TaoStats API

  • Price data via CoinGecko

References:

📊 Calculate your TAO Staking Rewards

⚙️ Access historic Bittensor Staking Data via API

PreviousBitcoin SRB
NextBNB Chain SRB

Last updated