Subnet 123: MANTIS
MANTIS is Bittensor Subnet 123. Its on-chain identity registers the subnet name as MANTIS and lists Barbariandev/MANTIS as the subnet’s GitHub repository. The repository describes MANTIS as a market-signal subnet in which miners publish encrypted forecast embeddings and validators score those submissions after the relevant market outcomes can be observed.
What MANTIS Provides
MANTIS turns financial forecasting into a set of recurring subnet challenges. The public configuration includes binary direction forecasts, ETH barrier-hit forecasts, volatility-regime and quantile-path forecasts, range-breakout predictions, cross-sectional return ranking, funding-rate ranking, and trade-mix signals. These tasks cover crypto assets, selected foreign-exchange pairs, and precious-metal exposure through challenge-specific input formats.
The subnet does not ask validators to judge free-form text or manual trade narratives. Instead, miners submit structured embeddings for each challenge. Validators later compare those embeddings with observed price or funding-rate data, then use challenge-specific scoring paths to estimate which miner signals were useful. The README describes the final score as per-hotkey salience that is aggregated across challenges before validators set weights on chain.
Challenge Diversity Context
The public README and miner guide show MANTIS as a multi-problem forecasting benchmark rather than a single market-direction contest. Its challenge set spans short-horizon direction calls, barrier-touch outcomes, volatility-regime distributions, range-breakout events, relative return rankings, and funding-rate rankings. Those problem families ask for different kinds of signal: some are binary, some are probability distributions, and others compare many assets against one another.
That diversity changes how the subnet should be read. A miner can be useful for one financial question without being broadly useful across every challenge, so the subnet’s final weight is not just a measure of whether a miner predicted one asset correctly. The README describes salience as computed inside each challenge and then aggregated by challenge weight. In practice, the public sources frame the subnet as a portfolio of scoring tasks where a miner’s contribution is judged across several market structures.
That framing also explains why the challenge names matter. The miner guide lists different horizons and output shapes for the challenge families, so a valid signal for one family is not automatically equivalent to a valid signal for another. The common embedding wrapper is a transport format for several distinct forecast problems, not a promise that every task is evaluated the same way.
The same sources also separate MANTIS from a copy-trading or trade-publication venue. Miners submit structured forecast embeddings, and validators evaluate those embeddings only after the relevant market outcome can be observed. The valuable artifact is therefore not a public trade instruction or a claim about an individual model’s private strategy. It is the evidence that a miner’s time-locked signal carried predictive value when compared with later market data.
For readers comparing MANTIS with other financial subnets, the important distinction is the breadth of labeled outcomes. The subnet uses crypto, foreign-exchange, metal, breakout, and funding-rate contexts as separate tests of market-signal quality. That makes its reward surface broader than a single price oracle, while still remaining narrower than a general finance assistant: the public materials describe recurring quantitative forecasting challenges, not discretionary research reports or portfolio management advice.
References: MANTIS README, MANTIS miner guide
Miner and Validator Roles
Miners generate model outputs for the configured challenges and publish them as encrypted V2 payloads. The miner guide describes a submission loop where a miner updates a Cloudflare R2 object whose object key matches the miner hotkey, while the on-chain commit points validators to that object. The payload format carries the hotkey, encryption metadata, ciphertext, and challenge embeddings.
Validators sample miner payloads, store raw submissions and challenge data in a SQLite ledger, and decrypt matured payloads using the Drand timelock path. The README describes a second owner decryption path, but validator scoring depends on the timelock path so predictions are not readable before maturation. After enough labeled data is available, validators run the model scoring code, smooth the resulting weights, and submit them to Bittensor. Those weights feed into Yuma Consensus, which converts validator agreement into miner and validator incentives.
On-Chain Identity
Live SN123 data is available on TaoStats. The live Finney identity for netuid 123 reports the subnet name as MANTIS and the GitHub repository as Barbariandev/MANTIS. The URL, contact, Discord, and description fields are currently blank.
Relationship to Yuma Consensus
Subnet 123 uses Yuma Consensus to convert the forecast-salience weight vectors that validators submit into the emission shares distributed to miners and validators within the subnet each tempo. The Yuma Consensus documentation describes how validator weight submissions are aggregated into consensus weights for each miner registered on the subnet.
In MANTIS’s context, validators decrypt time-locked forecast embeddings after market maturation, run challenge-specific scoring code to estimate per-hotkey signal salience, smooth the resulting weights, and submit those weight vectors to the subnet. The Emission documentation describes how those consensus weights determine each participant’s share of the subnet’s accumulated emission each tempo.
Development Stage Context
The Introduction to Bittensor describes subnet development as moving from localnet to testnet and then mainnet. For MANTIS (SN123), that sequence changes how readers should interpret market-signal embedding examples and challenge-based forecasting outcomes.
In localnet, MANTIS-compatible miners and validators can be developed and tested in an isolated environment. Localnet forecast embedding results and emission outcomes do not represent production subnet performance.
On testnet, MANTIS-compatible forecast submission and validator scoring workflows can be exercised in a shared, non-production network. Testnet salience scores and validator weights are separate from mainnet subnet state.
On mainnet, MANTIS (SN123) is the live production subnet where miners submit encrypted market-signal embeddings and validators score their salience across financial prediction challenges to determine real Bittensor emissions. The MANTIS repository is the registered project repository for SN123 on the production network.
The Bittensor Networks reference separates mainnet, testnet, and localnet. A forecast embedding result or emission outcome from one environment should not be read as representing production subnet performance in another environment.
Netuid 123 Identifies the Subnet On-Chain
Bittensor assigns every subnet a unique numeric identifier called a netuid, and Subnet 123 is the subnet registered at netuid 123 (Glossary: Netuid). The Understanding Subnets reference explains that each subnet runs its own incentive mechanism while sharing the same underlying Subtensor chain, so the netuid is the stable handle that distinguishes MANTIS from every other subnet.
For a reader, this means “Subnet 123” and “netuid 123” refer to the same on-chain slot. A claim about MANTIS should be tied to that netuid rather than to the registered name alone, because the name field can be changed on-chain while the netuid stays fixed.
Reader Boundary
Subnet 123 MANTIS should not be read as generic Bittensor subnet documentation, financial, trading, or investment advice, or a substitute for the subnet’s own primary sources. It names the on-chain subnet registered at netuid 123 under the identity “MANTIS” (Understanding Subnets, Glossary: Netuid).