How to Mine Vela (VELA)

Start mining Vela in minutes. Vela uses Proof-of-Inference v2: every share is a real sd-turbo-int neural-network inference on your NVIDIA GPU, mined with supr-meow-vela, our open-source C++/CUDA pool miner (MIT, no miner dev fee) for Linux, Windows, HiveOS/MMPOS/SMOS and Docker. Follow the 3 steps below to build your command, pick a port and connect. 1% FEE PPLNS PAYOUTS EVERY 15 MIN NVIDIA GPU

Miners online

group
Workers online:

Blockchain height

database
Network Difficulty:

Pool Hashrate

bar_chart
Nethash:
rocket_launch
Get mining in 3 steps
1 Your Vela address

Paste the Vela wallet address (starts with vela1) you want rewards paid to. The worker name (e.g. rig01) just labels this machine in Your Stats. No registration needed.

2 Platform, region & port

3 Connection & GPUs

Big rig? Add --suggest-diff 8192 (Docker/HiveOS: SUGGEST_DIFF=8192) on 7601 so VarDiff settles instantly. Leave GPUs empty to use every NVIDIA card in the box.

Your connect command (supr-meow-vela)

Download supr-meow-vela 0.1.0: Linux x86-64 · Windows x64 · HiveOS package · GitHub release & checksums · mirror on vela.suprnova.cc/dl · model sd-turbo-int.poisd (1.2 GB) — every launcher (HiveOS/MMPOS/SMOS, Windows start.cmd, Docker) downloads the model automatically; the bare CLI needs it next to the binary or via --model. Full per-platform instructions are below.

dns
Stratum Servers
Mining Pool Ping explained
Region Host Ports (plain TCP) TLS Ping
DEEU · Germany (main) vela.suprnova.cc 7601 · 7602 · 7603 7604 Getting Ping
UAEU 2 · Kyiv stratum-eu2.suprnova.cc 7601 · 7602 · 7603 7604 Getting Ping
USUS · Chicago stratum-ord.suprnova.cc 7601 · 7602 · 7603 7604 Getting Ping
HKAsia · Hong Kong stratum-hkg.suprnova.cc 7601 · 7602 · 7603 7604 Getting Ping

7601 VarDiff (start 2048, auto) · 7602 fixed 2048 · 7603 fixed 16384 · 7604 TLS. 1 difficulty = 1 inference; one share ≈ 20–30 s per RTX 5090 on 7601. Request a starting difficulty with --suggest-diff N (the password field is ignored). All regions feed the same pool (one PPLNS ledger, same address, same stats): pick the lowest ping. Regional hosts forward to Germany over WireGuard; the pool sees your real IP.

help
What goes where
Username: Your Vela wallet address (vela1…)
Workername: Add .workername to your Username
Algo: Proof-of-Inference v2 (sd-turbo-int) — built into supr-meow-vela
Password: x (ignored — use --suggest-diff N for a start difficulty)
Example: vela1q….myWorker01
terminal
Miner setup & downloads
github.com/ocminer/supr-meow-vela · v0.1.0 · MIT
GPU Miner
  • supr-meow-vela 0.1.0 GPU Recommended 0% miner fee · MIT
    Download: supr-meow-vela-0.1.0.tar.gz (Linux x86-64, bundled CUDA 13.3 runtime, NVIDIA driver R580+) · mirror · SHA256SUMS · source Model file: sd-turbo-int.poisd (1.2 GB, downloaded once; sha256 ba38ab96b3398e22…, identical to the Vela reference release) Open-source C++/CUDA Proof-of-Inference miner: exact int8 sd-turbo-int inference, bit-identical to the Vela reference verifier, several inferences in flight per GPU. One binary for Ampere (sm_80/86: CMP 170HX, A100, RTX 30xx, A6000), Ada (sm_89: RTX 40xx, 6000 Ada), Hopper (sm_90: H100/H200), Blackwell datacenter (sm_100: B200) and Blackwell consumer/pro (sm_120: RTX 50xx, PRO 6000). Volta/V100 is not supported. Install and run: wget https://github.com/ocminer/supr-meow-vela/releases/download/v0.1.0/supr-meow-vela-0.1.0.tar.gz tar xzf supr-meow-vela-0.1.0.tar.gz && cd supr-meow-vela-0.1.0 wget https://vela.suprnova.cc/dl/sd-turbo-int.poisd ./supr-meow-vela -o vela.suprnova.cc:7601 -u YOUR_VELA_ADDRESS.rig01 -p x --model sd-turbo-int.poisd With TLS/SSL: ./supr-meow-vela -o vela.suprnova.cc:7604 --tls -u YOUR_VELA_ADDRESS.rig01 -p x --model sd-turbo-int.poisd Tuning: --devices 0,1 selects cards, --batch 4 --lanes 2 are the defaults (try --batch 8 on 24 GB+ cards), --suggest-diff 8192 for big rigs. Measured: RTX 5090 / PRO 6000 ≈ 125–135 inferences/s, B200 ≈ 113, H200 ≈ 92, RTX 4090 ≈ 51, CMP 170HX ≈ 31. ./supr-meow-vela --check runs the spec test vector, --bench 20 a throughput test.
GPU Miner
  • supr-meow-vela 0.1.0 GPU Recommended native x64
    Download: supr-meow-vela-0.1.0-windows-x86_64.zip (Windows 10/11 x64, CUDA DLLs included, NVIDIA driver R580+) · mirror · SHA256SUMS Native Windows build — no WSL, no Docker. Extract the complete ZIP to a writable folder, open start.cmd in Notepad and set WALLET to your Vela address (optionally WORKER and DEVICES), then double-click start.cmd. The 1.2 GB model downloads on first start and is cached in the models folder. Or from PowerShell: .\start.ps1 -Wallet YOUR_VELA_ADDRESS -Worker rig01 -Devices 0 .\start.ps1 -Wallet YOUR_VELA_ADDRESS -Devices '0,1' -Batch 8 -Lanes 2 # two big GPUs .\start.ps1 -Wallet YOUR_VELA_ADDRESS -Pool vela.suprnova.cc:7603 # fixed-16384 port for large rigs The Windows build has no TLS: use the plain-TCP ports 7601/7602/7603. If Windows reports a missing VCRUNTIME/MSVCP DLL, install the Microsoft Visual C++ 2015-2022 x64 runtime. Expect the same rate as on Linux (RTX 5090 ≈ 125 inferences/s).
Cloud GPUs & hashrate rentals
  • vast.ai / RunPod / clore.ai / OctaSpace / any GPU host Works today
    Easiest: rent the instance with the Docker image ocminersupr/supr-meow-vela:0.1.0 and the environment variables WALLET and WORKER (see the Docker tab — vast.ai templates, clore.ai "custom image" and OctaSpace all take an image name plus env vars). Any instance with an NVIDIA Ampere-or-newer GPU where you can run your own binary also works with the Linux tarball: wget https://github.com/ocminer/supr-meow-vela/releases/download/v0.1.0/supr-meow-vela-0.1.0.tar.gz https://vela.suprnova.cc/dl/sd-turbo-int.poisd tar xzf supr-meow-vela-0.1.0.tar.gz && cd supr-meow-vela-0.1.0 nohup ./supr-meow-vela -o vela.suprnova.cc:7601 -u YOUR_VELA_ADDRESS.cloud01 -p x --model ../sd-turbo-int.poisd --suggest-diff 8192 > miner.log 2>&1 & Multi-GPU instances: all cards are used automatically; add --suggest-diff 8192 (Docker: -e SUGGEST_DIFF=8192) or use port 7603 so VarDiff does not have to ramp up first.
  • Nicehash / Mining Rig Rentals
    Proof-of-Inference is not a Nicehash or MRR algorithm — every share is a full sd-turbo-int inference on the rig's own GPU, so rented hashrate must run supr-meow-vela itself (see above). We will post a guide here if a marketplace adds the algorithm.
HiveOS Flight Sheet Setup
  • Step 1: Create a Flight Sheet
    Go to Flight Sheets in your HiveOS dashboard and click Add Flight Sheet. Set the following:
    CoinSelect Custom, enter VELA
    WalletCreate a new wallet with your Vela address (vela1…), or select an existing one
    PoolSelect Configure in miner
    MinerAdd supr-meow-vela as a custom miner (see Step 2)
  • Step 2: supr-meow-vela (Custom Miner) Recommended
    supr-meow-vela is a Custom Miner in HiveOS. Add it once with the installation URL below; the model file (1.2 GB) is downloaded on first start and cached in /hive-config/models.
    MinerSelect Custom at the bottom of the miner dropdown
    Miner namesupr-meow-vela
    Installation URLhttps://vela.suprnova.cc/dl/supr-meow-vela-hiveos.tar.gz (or https://github.com/ocminer/supr-meow-vela/releases/download/v0.1.0/supr-meow-vela-hiveos.tar.gz)
    Hash algorithmvela (label only)
    Wallet and worker%WAL%.%WORKER_NAME%
    Pool URLvela.suprnova.cc:7601
    Passwordx
    Extra config argumentsoptional, one per line: BATCH=8 · LANES=2 · DEVICES=0,1 · SUGGEST_DIFF=8192 · TLS=1 (then pool port 7604) · MODEL_URL=https://vela.suprnova.cc/dl/sd-turbo-int.poisd (default)
  • Step 3: Apply and Start
    1. Save the flight sheet and apply it to your worker(s). 2. The miner will start automatically. Check the Miner log tab to confirm shares are being accepted. 3. Your stats will appear on the Your Stats page within a few minutes. Tip: the first start takes a minute longer while the 1.2 GB model downloads. Multi-GPU rigs: add SUGGEST_DIFF=8192 so VarDiff settles immediately.
MMPOS and SMOS
  • MMPOS / SMOS setup
    The Linux tarball supr-meow-vela-0.1.0.tar.gz contains the launchers mmpos-run.sh (MMPOS) and smos-run.sh (SMOS) next to the binary. Add supr-meow-vela as a custom miner with that tarball, choose the launcher for your OS, and set these variables in the miner config: Pool: POOL_URL=vela.suprnova.cc:7601 Wallet: WALLET=YOUR_VELA_ADDRESS.%WORKER% The model downloads automatically (MODEL_URL defaults to the pool mirror, checksum-verified). Optional: BATCH=8 LANES=2 DEVICES=0,1 SUGGEST_DIFF=8192 TLS=1.
Mine Vela with Docker (vast.ai / clore.ai / RunPod / OctaSpace / any NVIDIA host)
  • ocminersupr/supr-meow-vela:0.1.0 GPU Recommended
    Requires a Linux host with an NVIDIA driver (R580+) and the NVIDIA Container Toolkit. The image also exists as :latest. Keep the model volume so the 1.2 GB download happens once. Run (one GPU): docker run -d --name vela --gpus '"device=0"' --restart unless-stopped \ -v vela-models:/models \ -e WALLET='YOUR_VELA_ADDRESS' -e WORKER=rig01 \ ocminersupr/supr-meow-vela:0.1.0 POOL_URL defaults to vela.suprnova.cc:7601 (PASSWORD to x). All GPUs: --gpus all -e DEVICES=0,1,2. Big cards: -e BATCH=8 -e LANES=2. Large rigs: -e SUGGEST_DIFF=8192 or -e POOL_URL=vela.suprnova.cc:7603. TLS is not built into the image — use a plain-TCP port. View mining progress: docker logs -f vela
  • vast.ai / clore.ai / OctaSpace / RunPod rentals
    Choose a custom Docker image deployment, enter the image ocminersupr/supr-meow-vela:0.1.0 and the environment variables WALLET=YOUR_VELA_ADDRESS WORKER=cloud01 (plus DEVICES=0,1,… on multi-GPU instances). No SSH or manual setup — the container starts mining by itself. Headless rigs: add SSH_PUBKEY='ssh-ed25519 …' (or SSH_PASSWORD, optional SSH_PORT) to get a shell inside the container. MODEL_PATH points at an existing model copy; EXTRA_ARGS passes extra miner flags.
quiz

Vela (VELA) mining — Frequently Asked Questions

How do I mine Vela (VELA)?
Download supr-meow-vela and the sd-turbo-int.poisd model from vela.suprnova.cc/dl, then run ./supr-meow-vela -o vela.suprnova.cc:7601 -u YOUR_VELA_ADDRESS.rig01 -p x --model sd-turbo-int.poisd. Use port 7604 with --tls for an encrypted connection. No registration — your wallet address is your account.
What is Proof-of-Inference and what hardware do I need?
Vela's PoW (Proof-of-Inference v2) requires a real int8 inference of the sd-turbo-int model for every attempt; the block is valid when the SHA-256d of the header built from that inference meets the target. You need an NVIDIA GPU, Ampere or newer (RTX 30/40/50, CMP 170HX, A100/A6000, H100/H200, B200, 6000 Ada / PRO 6000). RTX 5090 or PRO 6000 ≈ 125–135 inferences/s, B200 ≈ 113, H200 ≈ 92, RTX 4090 ≈ 51, CMP 170HX ≈ 31. CPU mining is not practical.
Which miner should I use?
supr-meow-vela 0.1.0 — the pool\'s open-source C++/CUDA miner (MIT, no miner dev fee, github.com/ocminer/supr-meow-vela). Its output is bit-identical to the Vela reference verifier (the pool verifies every share on GPU), it runs several inferences in flight per card and supports multi-GPU, VarDiff and TLS (Linux). Packages: native Linux and Windows builds, HiveOS/MMPOS/SMOS launchers and the Docker image ocminersupr/supr-meow-vela.
What is the pool fee and how are rewards paid?
1 % pool fee. Rewards are shared PPLNS over the last 5 rounds, so you are paid your share of every block the pool finds. Payouts run automatically every 15 minutes to your mining address once your balance reaches 0.01 VELA (lower it in Your Stats) and the block has 101 confirmations (≈ 8.5 h). The 10 % devfund is taken by the chain itself before the reward reaches the pool.
Which stratum port should I use?
7601 (VarDiff, start 2048, auto-adjusting) is right for everyone from one card to a farm. 7602 is fixed at 2048, 7603 fixed at 16384 for large rigs, 7604 is 7601 over TLS. Big rigs can add --suggest-diff 8192 to skip the ramp-up.
Why is my rate shown in inferences/s instead of hashes/s?
On Vela one unit of difficulty equals one complete model inference, so there is no meaningful "hash" count. The pool, the miner and this site therefore show inferences per second; network difficulty is the expected number of inferences per block (≈ 9 M at the moment, i.e. about one block per 5 minutes network-wide).
How do I get a Vela wallet address?
Run the Vela node/wallet from the official repository and create a receiving address (getnewaddress, it starts with vela1). That address is your mining username and your payouts go straight to it.
Can I mine Vela on vast.ai, RunPod or other cloud GPUs?
Yes — any instance with an NVIDIA Ampere-or-newer GPU where you can run your own binary. Easiest is the Docker image ocminersupr/supr-meow-vela with WALLET and WORKER set (see the Docker tab); the Linux tarball works on any instance too. Nicehash / MRR do not offer the algorithm.