10% below market
We track the market weekly and price H100 GPU rental and the token API underneath it. B300 pre-sale is offered at market rate — next-gen Blackwell Ultra capacity, no discount.
All GPU rates per GPU-hour · USD · refreshed weekly · final quote tailored to region, config & term
Available today — rent flexibly
Start on-demand with a single GPU — no minimums, no lock-in, cancel anytime. Commit to a longer term only when you're ready to save more.
On-demand
- Billed hourly, no commitment
- Scale up or down anytime
- Dedicated bare metal
- 24/7 support included
1-month reserved
- 30-day commitment
- Guaranteed capacity
- Priority provisioning
- 24/7 support included
6-month reserved
- 6-month commitment
- Guaranteed capacity
- Named account engineer
- 24/7 support included
12-month reserved
- 12-month commitment
- Guaranteed capacity
- Named account engineer
- Custom network & storage
Pre-sale now open
Lock your B300 capacity and rate before public availability. With demand outpacing supply, pre-orders get priority allocation and a protected price.
- 288GB HBM3e — next-gen throughput per dollar
- Prepay to lock both your slot and your hourly rate
- Priority allocation ahead of general availability
- Custom quotes by cluster size and term
Flat per-token pricing
Serve Kimi K3, DeepSeek V4 Pro, and DeepSeek V4 Flash through an OpenAI-compatible API — one flat rate, 10% under DeepSeek's off-peak list prices. No minimums, zero data retention.
Pay-as-you-go
Per-token billing with no minimums. Start with free trial credits and scale as you go.
Volume discounts
Committed token volumes unlock deeper per-token rates. Contact sales for a custom quote.
OpenAI-compatible
Same API shape, same SDKs. Swap the base URL and you're live in minutes.
| Model | Context | Input /1M tokens | Output /1M tokens | Cached input /1M |
|---|---|---|---|---|
| Kimi K3 · vision + reasoning | 1M | $2.70 | $13.50 | $0.27 |
| DeepSeek V4 Pro · flagship | 1M | $0.59 | $1.78 | $0.02 |
| DeepSeek V4 Flash 0731 · low cost | 1M | $0.20 | $0.59 | $0.006 |
Flat per-token rates, 10% below official off-peak list prices (DeepSeek peak = 2× off-peak). Official off-peak reference (per 1M): Kimi K3 $3.00 in / $15.00 out · DeepSeek V4 Pro $0.66 in / $1.98 out · DeepSeek V4 Flash $0.22 in / $0.66 out. Cached-input rates apply to prompt-cache hits. Committed volumes are locked.
See the difference
| Vetu Link | Market reference | Hyperscalers | |
|---|---|---|---|
| H100 on-demand | $3.38 /hr | $3.75 /hr | $6.9–$12.3 /hr |
| H100 12-month | $2.03 /hr | $2.25 /hr | $1.90–$2.10 /hr* |
| B300 pre-sale | $7.85 /hr | $7.85 /hr | $18 /hr* |
| Egress / hidden fees | $0 | Varies | Often significant |
| Tenancy | Dedicated bare metal | Mixed | Virtualized / shared |
| Provisioning | Under 1 hour | Hours–days | Complex setup |
Market reference = enterprise-tier GPU cloud on-demand (Lambda-class providers), refreshed weekly. All rates illustrative — final quote on request. * Hyperscaler figures reflect instance-level (multi-GPU) pricing and typically exclude egress fees.
Common questions
How does the 10%-below-market pricing work?
We sample public on-demand and reserved rates across major GPU clouds weekly, compute the market average, and set each of our tiers 10% under it. Your quoted rate is locked for the term you choose.
Is my data safe on dedicated hardware?
Yes. Every deployment is isolated bare metal, encrypted in transit and at rest, with zero data retention and a contractual commitment that we never train on your data.
Can I start small and scale later?
Absolutely. Start on-demand with a single node, then lock longer terms as your workload stabilizes to drop your effective hourly rate.
How do B300 pre-orders work?
Tell us your cluster size and term. We'll quote a locked rate; a prepayment reserves your allocation ahead of general availability. Contact sales for details.
Get a quote in under one business day
Tell us your workload, region, and term — we'll come back with a locked rate.