Questions? Start with the FAQ and the docs — customers with an active server reach an engineer by ticket, 24/7 for incidents.
GPU Cloud HQDedicated, billed monthly
vs
Together AIInference platform with GPU clusters
Best known for serverless model APIs, with dedicated GPU clusters sold alongside them by the hour or by reservation. Below: their published rates read on 15 October 2026, what they add up to over a full month, and the cases where Together AI is simply the better choice.
Every model Together AI publishes that we also rent. Their rate is the standard on-demand tier — spot and preemptible tiers are cheaper and noted where they exist.
| GPU | Together AI published rate | One month, 24/7 | Our monthly | Difference |
|---|---|---|---|---|
| NVIDIA B300288 GB HBM3e | $9.99/h on demand · $4.99 preemptible | $7,293 | $3,790 | −48% |
| NVIDIA B200180 GB HBM3e | $8.19/h on demand · $7.99 reserved 3–30 days · $4.09 preemptible | $5,979 | $2,690 | −55% |
| NVIDIA H200141 GB HBM3e | $5.99/h on demand · $4.99 reserved 3–30 days · $2.99 preemptible | $4,373 | $1,690 | −61% |
| NVIDIA H100 SXM80 GB HBM3 | $3.99/h on demand · $3.69 reserved 3–30 days · $1.99 preemptible | $2,913 | $1,249 | −57% |
Our price includes the dedicated CPU, RAM and NVMe listed on the pricing table, 20 TB of outbound transfer and a 99.9% SLA. Read on 15 October 2026 from together.ai.
Both of these lists are real. Pick the column that describes your workload.
Choose Together AI if you want a model API and a cluster from the same vendor, and you value the inference stack more than the price per hour.
Choose us if the GPUs stay busy, you want the bill to be predictable, and you would rather not hand over an identity document to rent a computer.
| What to check | Together AI | GPU Cloud HQ |
|---|---|---|
| How you are billed | Per hour on demand; preemptible billed every one to two minutes | Flat monthly price, prepaid balance, invoices settle themselves |
| Commitment | Reserved tiers from 3 days to 181+ days, paid upfront and non-refundable | Rolling month, or 3/6/12 months for 5–15% off |
| Signing up & identity | Self-serve, prepaid from a $15 minimum | Self-serve, email address only — no KYC |
| How you pay | Card (prepaid cards refused) or US ACH transfer; a card must stay on file | Bitcoin, Ether, USDT (ERC-20 / TRC-20) — no KYC, no card |
| What you actually get | Bare-metal clusters with InfiniBand, managed Kubernetes or Slurm | Single-tenant bare metal, root access |
| Extras | Serverless inference and fine-tuning APIs on the same account | Console with balance, invoices and auto-renew |
| Network | No data-transfer fees in or out | 10–25 Gbps dedicated, 20 TB egress included |
For a machine that stays busy, yes: Together AI publishes $3.99/h on demand · $3.69 reserved 3–30 days · $1.99 preemptible, which is $2,913 for a 730-hour month, against $1,249 flat here for a dedicated H100 — about 57% less. Below roughly 313 hours of use a month, their hourly rate works out cheaper.
You want a model API and a cluster from the same vendor, and you value the inference stack more than the price per hour. If that is you, stay with them — we would rather say so than sell you the wrong thing.
No. Every price on this page is public on both sides. When you do order here, an email address is all we ask for: no ID document, no company registration, no card — the balance is funded in Bitcoin, Ether or USDT.
They were read on 15 October 2026 from Together AI's own pricing page (https://www.together.ai/pricing). Providers change prices without notice — the link is there so you can check today's figure yourself.
Every GPU is in stock, the price is public, and the first invoice settles from your balance in minutes.
