Miners vs Validators: The Two Roles Behind Every Subnet
Bittensor runs on two jobs pulling in opposite directions. Here is what each does and how they get paid.
Almost everything on Bittensor comes down to two roles. Miners produce; validators judge. The tension between them is what keeps the network honest.
Miners: the producers
A miner runs the model or service a subnet is built around and competes to give the best response to the subnet’s task. Depending on the subnet, that might be answering prompts, serving low-latency inference, training a model, scraping and structuring data, or forecasting an event. Miners are ranked against one another, so there is constant pressure to improve.
Validators: the judges
A validator repeatedly queries the miners, evaluates their output using the subnet’s incentive mechanism, and submits scores - weights - to the chain. Validators carry stake, either their own or delegated to them, and that stake gives their scores influence. The more stake behind a validator, the more its opinion counts.
How the money flows
Miner rewards are set by the validators’ consensus scores, run through Yuma Consensus: better work, judged consistently across validators, earns a larger share of the subnet’s miner emissions. Validators, in turn, earn dividends for evaluations that agree with the stake-weighted majority - and those dividends are shared with the delegators who staked behind them.
Which role is for you
Mining is hands-on. It usually means hardware, a competitive model or pipeline, and ongoing tuning to stay near the top of a subnet. Validating is capital- and reputation-led: it rewards stake, uptime and accurate, consistent scoring - and most people take part in it indirectly, by delegating TAO to a validator they trust rather than running one themselves.
Neither role exists without the other. Miners with no validators would be unaccountable; validators with no miners would have nothing to grade. Bittensor’s design simply makes it pay to do each job well.