Bittensor
Table of Contents

⚠ Disclaimer: This entry may be incomplete, out of date, or inaccurate. It is AI-maintained on a best-effort basis. Do not rely on it as a sole source — verify claims independently using the sources listed below.

Summary

Bittensor is a blockchain-coordinated protocol (native token: TAO) that rewards useful off-chain digital work — including AI inference and training — through an on-chain incentive mechanism rather than a central platform operator. The network is organized into independent “subnets,” each a self-contained market where miners compete to produce a specific service and validators continuously score their output; scores translate automatically into token emissions via a stake-weighted aggregation rule called Yuma Consensus. This validator-scoring loop is, functionally, an automated quality-based routing mechanism operating without a human dispatcher — the same problem this section’s other entries solve with posted prices (OpenRouter) or on-chain auctions (Akash). Subnet 64, branded Chutes, is the network’s flagship inference marketplace and the most concrete example of the mechanism applied to routing LLM API requests to GPU suppliers.

Key Facts

  • Founded: 2021, by the Opentensor Foundation
  • Type: Protocol / decentralized network (not a single company) — token-incentivized machine-intelligence marketplace
  • Status: Live mainnet; both co-founders stepped down from formal Opentensor Foundation leadership roles in February 2026 as part of a push toward fuller decentralization (see Notable Developments)
  • Native token: TAO; reported market cap approximately $3.4B as of Q1 2026 (per third-party analysis, unaudited)
  • Ecosystem revenue: reported $43M in Q1 2026 subnet-service revenue (~$172M annualized run-rate), the majority attributed to the Chutes inference subnet — self-reported/on-chain figures, not independently audited

What It Is / How It Works

Bittensor’s core structural idea is the subnet: an independently defined competition for a specific kind of digital work (LLM inference, image generation, model training, data scraping, and dozens of other categories), each identified by a netuid and run by its own registered participants. Within a subnet, miners perform the work and validators — who must stake TAO to participate — query miners, score their responses against subnet-specific criteria (accuracy, speed, or other subnet-defined metrics), and submit those scores on-chain as weights. Yuma Consensus, Bittensor’s chain-level aggregation algorithm, combines validator weights using stake-weighted voting (so validators backed by more staked TAO have more influence, discouraging low-stake collusion) and converts the result directly into TAO emission splits between miners and validators. No human dispatcher decides which miner serves which request or how much a miner is paid — the incentive mechanism does both continuously, epoch by epoch.

Chutes (subnet 64) is the concrete, buyer-facing product built on top of this mechanism: a developer API and web console, styled as a decentralized alternative to a hosted-model API, that accepts inference requests, schedules them onto miner-run GPU nodes (via containerized model deployments, with node requirements such as A100/H100 class hardware specified per model), and uses “adversarial validation” — having multiple miners answer the same query so outputs can be cross-checked — to catch a miner that isn’t actually running the model it claims to be. Buyers pay per query or per token, settleable in TAO or fiat; miner payouts flow through the subnet’s TAO emissions rather than a separate invoicing system. As of a Q1 2026 ecosystem revenue report, Chutes had processed more than 9.1 trillion tokens for over 400,000 users and was the first Bittensor subnet to cross $100M in cumulative inference volume — notably, an estimated 20–25% of Chutes’ daily inference volume is reported to route through OpenRouter, making Chutes a wholesale supplier into the very posted-price aggregation marketplace documented elsewhere in this section rather than a fully separate channel.

Other subnets extend the same discovery-and-routing pattern to adjacent markets: Targon (subnet 4) positions itself as an enterprise inference/compute marketplace (reported ~$10.4M annualized revenue, with AI-character-app company Dippy cited as a customer), and Templar (subnet 3) applies the same miner/validator incentive structure to distributed model training rather than inference (reported completion of a 72.7B-parameter “Covenant-72B” model trained across 70+ contributor nodes on commodity GPUs). None of these figures were independently audited in the sources reviewed for this entry.

Claim Verification

Claim: “$43M in Q1 2026 AI service revenue / $172M annualized run-rate” and Chutes’ “9.1 trillion tokens / 400,000+ users / first subnet past $100M cumulative volume”

Status: Unverified

Supporting sources:

Refuting / questioning sources:

  • None identified in this review; no independent third-party audit of these figures was found.

Summary: These are on-chain-derived figures reported by a crypto-industry analysis site rather than an audited financial disclosure — treat as directionally informative but unverified, consistent with this section’s general caution around self-reported DePIN/DeFi network metrics.

Notable Developments

  • 2026-02-13: Co-founders Jacob Steeves (“Const,” CEO) and Ala Shaabana (“ShibShib,” COO) announced they were stepping down from their formal Opentensor Foundation leadership positions to accelerate the network’s transition toward full decentralization, while stating they would remain active as developers/contributors without the same legal authority.
  • 2026-05 (report date): Third-party analysis reports Bittensor’s subnet ecosystem crossed $43M in Q1 2026 service revenue (~$172M annualized), with the Chutes inference subnet identified as the primary driver — see Claim Verification above.
  • 2021: Bittensor founded by the Opentensor Foundation (Jacob Steeves and Ala Shaabana); the TAO token was distributed without a traditional venture fundraise, favoring broad participation over a priced round.

Key People

Jacob Steeves (“Const”) — Co-founder; CEO of Opentensor Foundation until stepping down from the formal role in February 2026 (remains an active contributor/developer per his own public statement).

  • LinkedIn: not found — TODO: verify current profile

Ala Shaabana (“ShibShib”) — Co-founder; COO of Opentensor Foundation until stepping down from the formal role in February 2026.

  • LinkedIn: not found — TODO: verify current profile

People — Last Reviewed: 2026-08-15

Sources

← prev