All Subnets

grail

#81Training

Decentralized post-training reinforcement learning (RLHF/GRPO) subnet - the alignment and quality layer for Covenant AI's training stack.

Price

$0.026793+5.52%

24h

7d Change

+27.80%

Market Cap

$30.96M

Emission

6.55%

Miners

2

Validators

8

Competitor Mapping

grail

Decentralized

Market Cap

$30.96M

TAM

$30.00B

Model

Token-incentivized

Team

Decentralized miners

OpenAI (RLHF) + Anthropic (Constitutional AI)

Private

Market Cap

Private

TAM

$30.00B

Model

Revenue-driven

Team

Enterprise

Grail decentralizes the post-training RL step that transforms base models into aligned, capable systems - the same process that made GPT-4 dramatically better than GPT-3. Competes with OpenAI's RLHF and Anthropic's RLAIF pipelines.

Implied Valuation

If grail captures a share of OpenAI (RLHF) + Anthropic (Constitutional AI)'s $30.00B TAM

Bear - 0.1% TAM

$30.00M

implied valuation

-3.10%

vs $30.96M current

Base case

Base - 1% TAM

$300.00M

implied valuation

+869.02%

vs $30.96M current

Bull - 5% TAM

$1.50B

implied valuation

+4745.12%

vs $30.96M current

Scenarios illustrate potential scale relative to the traditional market. Actual outcomes depend on adoption, tokenomics, and competitive dynamics.

Thesis

Post-training RL is the most valuable step in the AI pipeline - it's what turns raw models into useful products. Grail completes Covenant AI's vertical stack (SN3 pre-training, SN39 compute, SN81 post-training), creating a full decentralized alternative to OpenAI's pipeline. The 211% 30-day token surge after Covenant-72B shows market conviction.

Team

Part of Covenant AI, founded by Sam Dare (same team as SN3 Templar). Covenant AI operates three subnets: Templar (SN3, pre-training), Basilica (SN39, compute), and Grail (SN81, post-training/RL). Team is doxxed.

  • Sam Dare - Founder & CEO (Covenant AI) - 8+ years emerging tech, also leads SN3 Templar

Funding

No confirmed VC funding rounds disclosed. Funded through TAO emissions alongside SN3 and SN39.

Traction

SN81 token surged 211% in 30 days (Mar 2026). Running GRPO-based reinforcement learning on validated rollouts. Miners generate rollouts, validators verify proofs and evaluate kernel correctness on-GPU. Pushing into kernel verification for provable training computation.

Recent News

  • Mar 2026 - SN81 token surged 211% in 30 days following Covenant-72B announcement
  • Mar 2026 - Grail v0 shipped decentralized reinforcement learning capability
  • 2026 - Working on kernel verification to prove training computations happened correctly
  • 2026 - Covenant AI vertical integration: SN3 + SN39 + SN81 forming 'decentralized OpenAI'

Risk

Entirely dependent on SN3 Templar producing base models worth fine-tuning. If Covenant-72B proves to be a one-time achievement rather than repeatable pipeline, Grail loses its input.