Curious about a word? Tap a dotted term for its meaning.

AI incentive blockchain / NETWORK PROFILE

Bittensor

TAO CMC #33 in the snapshot

Bittensor uses the Subtensor blockchain to coordinate markets for digital work, including inference, predictions and computation. The useful work occurs in subnets; the ledger accounts for participation and incentives.

Profile checked Asset ranks captured 14 Nov 2026Snapshot & methodology ↗

01 / THE PURPOSE

What is Bittensor for?

Each subnet defines a task and a scoring mechanism. Miners supply outputs, subnet validators evaluate them, and token incentives try to direct resources toward useful services. An attractive reward does not itself demonstrate that the output is accurate.

02 / THE ARCHITECTURE

Three questions about the machinery.

01

How does it agree?

Consensus & security

Current official documentation identifies Subtensor as proof of authority: approved authorities produce blocks with Aura and finalize them with GRANDPA. Yuma Consensus aggregates evaluations for incentives; it does not finalize the blockchain. A proposed move to nominated proof of stake remains a plan.

02

Where do instructions run?

Execution model

The Substrate-based chain processes registrations, staking operations, weights and transfers. Model training or inference generally runs outside the replicated ledger, so blockchain verification of accounting is different from independently verifying every AI answer.

03

What is the token for?

TAO in the system

TAO is the native asset. Subnets also have distinct alpha tokens paired with TAO in liquidity pools. Staking into a subnet exchanges TAO for its alpha; withdrawing reverses that exchange at the available pool price.

03 / THE DESIGN CHOICES

What comes with the design?

Each architecture makes choices about security, performance, coordination, and who can participate.

  • Subnet staking introduces alpha price movement, pool slippage and incentive-design risk.

  • Scoring can be gamed or reward narrow benchmarks; chain finality does not certify real-world usefulness.

04 / IN PRACTICE

Follow one use.

A translation subnet can reward answers that its evaluators score highly. A staker backs that subnet through its pool and may receive fewer TAO on exit even after earning alpha rewards.

05 / CHECK THE EXPLANATION

Sources & snapshot.

Network descriptions use the references below. The CoinMarketCap snapshot determines which assets are in scope and their captured rank; it does not assess the quality of a network.

Profile checked 14 Nov 2026. Network designs and ranked positions can change.