Rent a dedicated H100, H200, A100 or L40S by the month, paid in crypto with no KYC

A WeightsAPI dedicated GPU is reserved inference capacity for one model you choose. The operator deploys it, checks the fit and hands you the endpoint. From that activation date the machine is yours for one month with no per-token charge, so a busy month costs exactly what a quiet one does.

The rental ends one month after activation and never renews on its own, so a forgotten GPU cannot drain your balance.

Need root to run your own stack? A raw GPU server suits you better. Want a model served without operating the machine? Create a username account and order here.

A flat month that undercuts hourly billing for always-on work

An H100 SXM is $1,352 a month, or $1.85 an hour across 730 hours. RunPod lists its H100 SXM at $3.49 an hour and DeepInfra its H100 at $2.20, so a model serving traffic all day costs less here, and you know the bill in advance. See the monthly comparison.

One balance for tokens and GPUs

The credit you top up pays per-token requests on 15 open-weight models and prepays dedicated GPUs. Keep spiky or experimental work on per-token billing and move steady traffic to your own machine, with no second account or payment route. DeepInfra also sells both, but documents funding through Stripe Checkout. Browse the 15 models.

No identity check, and a clear record policy

Signup asks for a username (3 to 32 characters) and a password (at least 15): no email, phone, ID document or card, and nothing more at GPU checkout. RunPod tells crypto payers to complete any required KYC verification before their first payment. No KYC is not anonymity: orders, payments and usage are recorded against your account. See exactly what is asked and recorded.

Nine crypto options, no platform deposit fee

Top up in BTC, ETH, XMR, SOL, LTC, TRX, USDT on TRON or Ethereum, or USDC on Ethereum. The USD amount is fixed for 10 minutes, credited within about a minute once the network confirmations are reached, and you pay only the network fee, so a Monero-funded H100 month is an ordinary order. See every asset and confirmation count.

Your month starts at activation, not at payment

Provisioning never eats into the time you paid for. Cancel before activation and the rental payment returns to your WeightsAPI balance, ready for tokens or another GPU. Read the rental terms.

GPU memory must hold the weights plus the attention (KV) cache for your context and concurrency. The table shows weight size at 16-bit and 8-bit (billions of parameters × 2 or × 1 byte) and the smallest configuration with headroom. The operator confirms the final fit before activation.

DeepSeek V3 and R1 show why one balance matters: their FP8 weights exceed our largest 640 GB configuration, so they stay on per-token billing while a 24B to 70B model runs on your own GPU.

Hourly clouds bill only while a machine runs; we bill a flat month, so we multiplied each published hourly rate by 730 hours. H100 figures are SXM for WeightsAPI, RunPod and Lambda; DeepInfra and BTCServers list “H100”. Lambda’s rates exclude tax.

Where the flat month wins: any machine that runs around the clock. We are below RunPod, Lambda and DeepInfra on every configuration where we found a published rate; an 8 × H100 month is $9,736 here against about $20,382 at RunPod and $12,848 at DeepInfra.

Where hourly wins: short or bursty jobs. Our H100 month equals 387 hours at RunPod’s rate and 615 at DeepInfra’s; below that, pay by the hour. RunPod and Vast.ai bill per second, which suits batch jobs.

Where BTCServers wins: raw hardware price. It is also no-KYC and crypto-paid, and lists an H100 at $1,190 and an A100 80 GB at $690 a month as servers you operate yourself. Pick it if you want to run the machine; pick WeightsAPI for a model deployed for you and one balance that also pays per-token calls.

Total cost? The monthly price is the whole machine charge: no per-token fees, no platform deposit fee, no renewal. Add only the network fee. A top-up of $1,352 or more covers an H100 month; any remainder stays usable for tokens or a later order, though credit cannot be withdrawn to a wallet.

Why USD 20 per top-up? Every configuration costs more than that, so the minimum never blocks a rental, and balances below USD 20 stay usable after your first qualifying top-up. To test tokens with a few dollars, a micro-deposit aggregator fits better.

Is capacity guaranteed? Capacity is finite, so the operator checks hardware, region, model fit and timing for each order before activation, and your month starts only then. No standard delivery time is published: leave a buffer before a launch, and if the timing does not suit you, cancel before activation to move the payment back to your balance.

Payment pitfalls? Wait for confirmed credit, then pay the order from your balance. Send the exact amount on the network shown: late or different amounts go to review, and wrong-network transfers cannot be recovered.

What is recorded? Username, key hashes, usage, charges, payment details (asset, network, address, transaction) and GPU order details, including dates and endpoint. Full list on trust and security.

What can I rely on? There is no SLA, service credit or failover, so keep a fallback for critical traffic. What limits your exposure is the structure: the month starts at activation, cancelling before activation refunds your balance, and nothing auto-renews. Read the service conditions and your model’s license.

Choose a WeightsAPI dedicated GPU if one open-weight model serves traffic most of the day and you want a fixed monthly bill; you pay in XMR, BTC, USDT or another listed asset without KYC; you want the model deployed for you; or you want one balance for tokens and GPUs.

Look elsewhere if your GPU runs only part of the day (under about 387 hours a month against RunPod, 615 against DeepInfra); you need root, containers or training; you need an SLA, failover or autoscaling; you want the cheapest raw H100 or A100; or you need a machine within hours.

What to do now: create a username account and store the recovery code offline, pick a configuration from the fit table, top up from USD 20 (see billing), then place your order; your month starts only on the activation date the operator confirms.