Log change over the last 30 days +171.27%: protocol buys +0.12%, root sells −35.16%, users net +10.29%, cross-subnet +196.03%, unattributed +0.0004%. 4 days still in progress are counted as far as known.

Protocol buy
User buy (staking)
Cross-subnet
Root sell (dividend claims)
User sell (unstaking)
Residual
Price path
Market candle

About this chart

The chart follows the alpha price in τ. Each period starts at its opening price and steps through the mechanisms below, in a fixed order, to its closing price: a step that rises pushed the price up, one that falls pushed it down, and its height is how far it moved the price. A thin line carries each close into the next period’s open; it is dashed where the two differ. The faint candle behind each period is the market’s own open, high, low and close, for context.

With many periods on screen (90 days, or hours), the steps would be too thin to read, so the chart draws the price as a line instead and colours each period by the mechanism that moved it most. Hover a period for the full step-by-step breakdown.

Protocol buy: The chain spending excess emission TAO to buy alpha from the pool. A permanent protocol buyback; it pushes the price up.
Root sell: Root-dividend claims sold into the pool; they push the price down.
User buy / sell: Staking into and unstaking from this subnet.
Cross-subnet: Swaps from other subnets routed through this pool.
Residual: What the decomposition cannot attribute to a mechanism. It is always shown, never folded into another part.

Not here, because they do not move the spot price: the per-block TAO injection into the pool, and burns.

All figures are log changes, not simple percentage changes, so the parts add up exactly to the total; on big moves the two differ (a 10% log fall is a 9.5% price fall).

Today so far

2026-11-10 UTC, as of ≈00:25 UTC (block 9,472,216)

Price, today so far: 0.02492 τ
Volume, today so far: 23.01 τ
Market cap incl. burned, today so far: 7.19K τ
TAO in pool, today so far: 4.16K τ
Emission share, today so far: —not available yet

The day isn’t over yet, so these figures keep changing. They are in no daily total, average or rank.

Subnet Configuration

Pool & Mechanism

AMM Pool

Balancer

Mechanism

dTAO

Access

Registration

Open

Neurons

256/256Competitive

Validators

4/6460 open

Economics

Owner Cut

18.00%

Weights

Commit-Reveal

Enabled (v3, 1 epochs)

Liquid Alpha

Disabled

Weights Version

Weights Range

1–65535

Timing

Tempo

360~1.2h

Immunity Period

7,200~1.0d

Activity Cutoff

5,000~16.7h

Market & Economics

Market

Market Cap

6,887.7661 τ

FDV

11.36K τ

TVL

8,327.6817 τ

Lifetime Volume

241.33K τ

Holders

756

14d price

Revenue / Recycle

46.88×

last day · Net positive

Emissions

1.04%of network emissions

Emission history is being recalculated from on-chain allocation.

TAO emitted to this subnet

Sent into its pool each day: its share of network emissions

37.4003 τ/day

37.4003 τ per day

Daily distribution

Alpha paid out each day, valued at the pool price

7,200.0000 र/day

≈ 179.4500 τ·179.4500 τ per day

Miners58.91%

4,241.1926 र ≈ 105.7058 τ/day

Validators23.10%

1,662.8404 र ≈ 41.4440 τ/day

Owner18.00%

1,295.9670 र ≈ 32.3002 τ/day

Stake 1,000 τ here

399.3525 र/day≈ 9.9533 τ·9.9533 τ per day

From the validator share, before commission

~363.30%staker APY (est.)

· 1,573.05% total subnet

Pool emits 1.0000 र/block at 0.02492 τ/र. Yield varies with stake weight, pool price, and emission rate.

For Miners

58.91% of emission

Slots

256/256Competitive

Burn Cost

0.5000 τ

Est. per Miner

1.0146 र/epoch≈ 0.0253 τ

Pool / Epoch

212.0596 र≈ 5.2853 τ

For Validators & Stakers

23.10% of emission

Validator Slots

4/6460 open

Est. per Validator

20.7855 र/epoch≈ 0.5180 τ

Registration Activity

Neuron registration trends

This Interval

24h: 2 registrations (2 burn · 0 PoW)

Target / Interval

Slots Available

Competitive

256/256

Neurons

256

Development Activity

reaxlabs/reax-subnetPython

132 contributors97 commits/30dLast commit 5d ago

MIT

Top Contributors

fstandhartinger1

Ihor-524661

Recent Releases

v1.2.0-rc11.post8.1-kit111d agopre v1.2.0-rc11.post8.113d agopre v1.2.0-rc11.post813d agopre v1.2.0-rc11.post714d agopre v1.2.0-rc1118d agopre

What REAX does

Machine learning inference systems deployed across multiple operators face a fundamental challenge: verifying that remote inference nodes are actually running the claimed model with the correct weights and parameters, not substituting cheaper approximations or quantized alternatives. In centralized systems, this is solved through direct control; in decentralized networks, it requires cryptographic proof or expensive redundant computation.

REAX addresses this by creating a network where validators can verify miner answers through independent recomputation. Rather than trusting a miner's claim about what model it runs, validators recompute the same decision on their own copy of the pinned reference model and compare the numerical outputs (logit-space distances). If the miner's answer diverges beyond calibrated tolerances, the validator catches it immediately. This approach is incentive-compatible: cheating by running cheaper quantized models or alternative models produces measurably different logits that validators detect within rounds, leading to strikes that reset earnings.

REAX implements this on Bittensor as a subnet (SN92) where miners run pinned reference models (Plumb-4B or Surogate Rune 26B) in Docker sidecars on loopback, answering signed requests over HTTP. Validators run their own sidecars and verify a subset of each response against their reference runtime. The protocol pins model weights by hash, runtime by digest, prompt and temperature exactly, and uses logit-space distance as a deterministic, model-agnostic proof of answer fidelity. Capacity probes measure throughput simultaneously across all miners, preventing sybil attacks where multiple hotkeys on one machine claim independent capacity.

Auto-generated from the GitHub repository and may not reflect the latest changes.

About

active

REAX is a Bittensor subnet where miners serve pinned reference models (Plumb-4B for s1-fast, Surogate Rune 26B-A4B v3 for s1-pro) in a runtime sidecar and validators verify answers by recomputing them against their own reference runtime, comparing results in logit space to ensure exact model reproduction.

How it works

REAX uses a two-process architecture per hotkey. (1) A runtime sidecar runs the pinned REAX reference model (Plumb-4B for s1-fast or Rune 26B for s1-pro) in a Docker container on loopback (127.0.0.1:8101 or :8102) and serves an internal decision contract (POST /internal/v1/decide) with an HMAC token, never published. (2) The miner process (reax-miner) talks to the sidecar over loopback with the shared token, checks the hotkey is registered, publishes axon and commitment on chain, and serves signed requests (POST /v1/s1/decide) over the public protocol reax-s1/1. Validators run their own reference sidecars (one per enabled pool) on their GPU and recompute fresh items to verify miner answers. The validator never opens a port; it probes miners' public endpoints and scores based on logit-space distance and latency.

Multinode option (REAX_FEATURE_MULTINODE) allows one hotkey to receive decisions from several worker machines via reax-runtime-relay over TLS/VPN with HMAC-per-request authentication; the gateway (reax-miner) holds the hotkey and signs all responses.

Incentive mechanism

The validator generates 8 fresh typed decisions per miner per round from a cryptographic seed (no answer key can leak). After receiving answers, it draws a verification set (8 items for s1-fast, 3 for s1-pro) and recomputes them on its reference runtime. Per verified item, it computes logit-space distance d as the largest difference between miner and reference option logits (probabilities floored at 1e-6, invariant to temperature).

The mean distance D over verified items determines scoring: D ≤ 0.20 scores (0.8 + 0.2×speed_factor); D > 0.20 scores 0; D > 0.50 or any d > 2.0 causes a strike resetting EMA to 0. Speed scores linearly from fast bound (0.5s for s1-fast, 1s for s1-pro) to slow bound (8s, 12s) over 0.5 rounds; beyond slow bound request times out. Per-pool capacity probes run approximately every 10 rounds, load every UID simultaneously, and measure verified throughput.

Pay within a pool is ema × min(cap, measured_capacity / target_dps). Pools share: s1-fast 0.35, s1-pro 0.65 of inference part. Competition and specialist tracks have separate budgets and winner-take-most sharing (75% rank 1, 17% rank 2, 8% rank 3).

The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository (github.com, opens in a new tab, external site)

On-Chain Identitysigned on-chain by subnet owner

र

REAX

Ra · Devanagari letter · pronounced “ra”

Verified decision-model inference on Bittensor.

Owner: 5GMxp…mU6B
Hotkey: 5H3xd…ZXvt
Owner Cut: 18.00%
Contact: [email protected]

Registered: 1 month ago(Oct 9, 2026)
Block: 9,247,819
Slot history: Every subnet in SN92
Neurons: 256/256
Tempo: 360 blocks

X: https://x.com/reaxlabs (x.com, opens in a new tab, external site)

Website

URL: https://reax.co/ (reax.co, opens in a new tab, external site)
Status: Up200
TLS: Valid
Title: REAX — Decisions at reflex speed
Checked: 2026-11-09 04:06 UTC

Changes (10)

2026-11-07 04:06 UTCUp200REAX — Decisions at reflex speed
2026-11-05 04:05 UTCUp200REAX — Decisions at reflex speed
2026-11-04 04:04 UTCUp200REAX — Decisions at reflex speed
2026-11-03 04:08 UTCUp200REAX — Decisions at reflex speed
2026-11-01 04:06 UTCUp200REAX — Decisions at reflex speed
2026-10-29 04:05 UTCUp200REAX — Decisions at reflex speed
2026-10-28 04:06 UTCUp200REAX — Decisions at reflex speed
2026-10-23 04:06 UTCUp200REAX — Decisions at reflex speed
2026-10-22 04:06 UTCUp200REAX — Decisions at reflex speed
2026-10-21 04:06 UTCUp200REAX — Decisions at reflex speed

Subnet History

SubnetOwnerHotkeySet 2026-10-15 13:18 UTC

Block 9,288,972Extrinsic

SubnetIdentitySet 2026-10-14 17:28 UTC

Block 9,283,023Extrinsic

Emissions started 2026-10-14 08:25 UTC

Block 9,280,308Extrinsic

Subnet registered 2026-10-09 20:07 UTC

Block 9,247,819

SubnetOwnerHotkeySet 2026-10-09 20:07 UTC

Block 9,247,819

Recent Activity

SubnetOwnerHotkeySet Block 9,288,972 Extrinsic2026-10-15 13:18 UTC
SubnetIdentitySet Block 9,283,023 Extrinsic2026-10-14 17:28 UTC
FirstEmissionBlockNumberSet Block 9,280,308 Extrinsic2026-10-14 08:25 UTC
NetworkAdded Block 9,247,8192026-10-09 20:07 UTC
SubnetOwnerHotkeySet Block 9,247,8192026-10-09 20:07 UTC

This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository (github.com, opens in a new tab, external site) for the latest instructions.

Hardware Requirements

Miner

GPU ModelNVIDIA GPU (RTX PRO 6000 Blackwell, H100, RTX 4090, RTX 5090, RTX 3090, L40S, RTX A6000)

RAM32 GB

CPU Cores4

OSLinux x86_64

NetworkPublic IPv4/IPv6 address with one open TCP port, NTP required

Not yet verified: GPU, VRAM, GPU count, Storage, Notes

Validator

GPURequired

GPU ModelNVIDIA GPU 80-96 GB class (RTX PRO 6000 96 GB, H100 80 GB)

VRAM80 GB

OSLinux x86_64

NetworkOutbound only to miners' public ports and subtensor endpoint

Not yet verified: GPU count, RAM, Storage, CPU cores, Notes

Machine-extracted from the subnet's repository docs; may be wrong or out of date. Check the repository before buying or renting hardware. Open the repository (github.com, opens in a new tab, external site) · Extracted 9 Nov 2026

Miner Setup

Install bittensor CLI in separate virtual environment for wallet management

pip install bittensor-cli==9.23.2

Create coldkey and hotkey (coldkey stays offline, only copy hotkey to server)

btcli wallet new_coldkey --wallet.name reax && btcli wallet new_hotkey --wallet.name reax --wallet.hotkey miner1

btcli subnets register --netuid 92 --wallet.name reax --wallet.hotkey miner1

Download s1-fast model weights (public, ~9 GB)

hf download crh225/plumb-4b --revision 55de037801a8a9b9de3db5c0e16cef86210c2186 --local-dir /srv/reax/plumb-4b

Build REAX miner Docker image from repository root

docker build -t reax:1.2.0rc3 .

Build s1-fast runtime sidecar image

docker build -f runtimes/s1-fast/Dockerfile -t reax-s1-fast-runtime .

Create shared internal token between sidecar and miner (0600 mode required)

mkdir -p deploy/secrets && openssl rand -hex 32 > deploy/secrets/runtime-token && chmod 0600 deploy/secrets/runtime-token

Copy example miner config and fill in netuid, wallet names, public IP, mechanism/model

cp deploy/miner-s1.env.example deploy/miner.env

Start sidecar and miner with docker compose (both on host networking, loopback only)

REAX_UID=$(id -u) REAX_GID=$(id -g) REAX_S1_FAST_WEIGHTS=/srv/reax/plumb-4b docker compose -f deploy/docker-compose.s1.yml up -d --build

Run self-check against published reference answers (must pass before registration)

REAX_UID=$(id -u) REAX_GID=$(id -g) REAX_S1_FAST_WEIGHTS=/srv/reax/plumb-4b docker compose -f deploy/docker-compose.s1.yml run --rm --no-deps --entrypoint reax-selfcheck miner

Validator Setup

Install bittensor CLI in separate virtual environment

pip install bittensor-cli==9.23.2

btcli subnets register --netuid 92 --wallet.name reax --wallet.hotkey validator

Stake at least 1000 TAO and wait one tempo for validator permit

btcli stake add --netuid 92 --wallet.name reax --wallet.hotkey validator --amount <TAO>

Build REAX validator Docker image from repository root

docker build -t reax:1.2.0rc3 .

Build s1-fast reference runtime sidecar

docker build -f runtimes/s1-fast/Dockerfile -t reax-s1-fast-runtime .

Create shared internal token for reference sidecar

mkdir -p deploy/secrets && openssl rand -hex 32 > deploy/secrets/runtime-token && chmod 0600 deploy/secrets/runtime-token

Download s1-fast model weights (~9 GB)

hf download crh225/plumb-4b --revision 55de037801a8a9b9de3db5c0e16cef86210c2186 --local-dir /srv/reax/plumb-4b

Start s1-fast reference sidecar on loopback 127.0.0.1:8101

docker run -d --name reax-s1-fast-ref --network host --gpus all --init --restart unless-stopped --user "$(id -u):$(id -g)" -e S1M_INTERNAL_TOKEN_FILE=/run/s1m-internal-token -v "$PWD/deploy/secrets/runtime-token:/run/s1m-internal-token:ro" -v /srv/reax/plumb-4b:/model:ro reax-s1-fast-runtime

Copy example validator config

cp deploy/validator.env.example deploy/validator.env

Start validator process (opens no port, needs no GPU itself)

docker run -d --name reax-validator --network host --init --restart unless-stopped --user "$(id -u):$(id -g)" --no-healthcheck --env-file deploy/validator.env --entrypoint reax-validator -e REAX_MECHANISM=s1 -e REAX_MODEL=s1-fast -e REAX_S1_FAST_RUNTIME_URL=http://127.0.0.1:8101 -e REAX_RUNTIME_TOKEN_FILE=/run/reax-runtime-token -v $HOME/.bittensor/wallets:/wallets:ro -v "$PWD/deploy/secrets/runtime-token:/run/reax-runtime-token:ro" -v reax-state:/state reax:1.2.0rc3

Use shadow mode for first hour to let miners build non-zero average before weights submission begins

Run with REAX_SET_WEIGHTS=0 for first hour in shadow mode

No hardware? Back this subnet by staking instead New to mining? Start with the setup guide Validator setup guide

What moved the price

Overview

Today so far

Subnet Configuration

Market & Economics

For Miners

For Validators & Stakers

Registration Activity

Development Activity

What REAX does

About

Website

Subnet History

Recent Activity

Participate

Hardware Requirements

Miner Setup

Validator Setup

What moved the price

Today so far

Subnet Configuration

Market & Economics

For Miners

For Validators & Stakers

What REAX does

About

Website

Subnet History

Hardware Requirements

Miner Setup

Validator Setup