Two ways to run GPU workloads
Pay per second for experiments. Flat monthly rate for production. No hidden fees.
Serverless Jobs
Pay only for execution time. Jobs scale to zero automatically when idle — no charges while your GPU waits.
- Per-second billing, no minimums
- Scale to zero automatically
- Cold start ~60 seconds
- Best for: experiments, batch jobs, CI/ML pipelines
A100 80GB — $2.36/hr · $0.000656/sec
Persistent Endpoints
ProductionDedicated GPU, always on. Flat monthly rate — up to 20% cheaper than running Jobs 24/7.
- Zero cold-start latency (always warm)
- Fixed monthly billing — no surprises
- Custom domain & dedicated SLA
- Best for: production APIs, real-time inference
A100 80GB — $1,400/mo · $1.94/hr effective · 18% vs Jobs
Serverless Jobs pricing
Per-second billing. Scale to zero. No minimum usage.
| GPU | VRAM | Per Hour | Per Second |
|---|---|---|---|
| L4 24GB | 24 GB | $0.66 | $0.000183 |
| RTX 4090 24GB | 24 GB | $1.00 | $0.000278 |
| L40S 48GB | 48 GB | $1.46 | $0.000406 |
| A100 80GBPopular | 80 GB | $2.36 | $0.000656 |
| H100 PCIe 80GB | 80 GB | $3.59 | $0.000997 |
| H100 SXM 80GB | 80 GB | $3.77 | $0.001047 |
| H200 141GB | 141 GB | $4.31 | $0.001197 |
Persistent Endpoints pricing
Dedicated GPU reserved exclusively for you. Always-on, always warm.
A100 20GB
20 GB VRAM
$450/month
≈ $0.63/hr effective for 24/7
28% cheaper than Jobs 24/7
vs $0.87/hr × 720 hrs = $626/mo on-demand
- Zero cold-start latency
- Dedicated GPU — not shared
- Custom domain included
- 99.9% uptime SLA
RTX 4090
24 GB VRAM
$600/month
≈ $0.83/hr effective for 24/7
17% cheaper than Jobs 24/7
vs $1.00/hr × 720 hrs = $720/mo on-demand
- Zero cold-start latency
- Dedicated GPU — not shared
- Custom domain included
- 99.9% uptime SLA
A100 40GB
40 GB VRAM
$890/month
≈ $1.23/hr effective for 24/7
28% cheaper than Jobs 24/7
vs $1.71/hr × 720 hrs = $1,231/mo on-demand
- Zero cold-start latency
- Dedicated GPU — not shared
- Custom domain included
- 99.9% uptime SLA
Most Popular
A100 80GB
80 GB VRAM
$1,400/month
≈ $1.94/hr effective for 24/7
18% cheaper than Jobs 24/7
vs $2.36/hr × 720 hrs = $1,699/mo on-demand
- Zero cold-start latency
- Dedicated GPU — not shared
- Custom domain included
- 99.9% uptime SLA
H100 SXM 80GB
80 GB VRAM
$2,500/month
≈ $3.47/hr effective for 24/7
8% cheaper than Jobs 24/7
vs $3.77/hr × 720 hrs = $2,714/mo on-demand
- Zero cold-start latency
- Dedicated GPU — not shared
- Custom domain included
- 99.9% uptime SLA
Need a different GPU config? Endpoints are available for all GPU types. Contact sales@velar.run for H100, H200, multi-GPU setups, or custom monthly pricing.
Platform plans
Choose a plan that matches your scale. GPU costs are billed separately.
Free
$0/month
Get started with $10 in free credits
$10 one-time credits included
Get Started Free- Serverless Jobs only
- Up to 2 concurrent GPUs
- Per-second billing
- Community support
- Basic monitoring
- Scale to zero
Most Popular
Pro
$49/month
For developers scaling AI workloads to production
$20 credits/month included
Start Pro Trial- Serverless Jobs + Persistent Endpoints
- Up to 10 concurrent GPUs
- 1 Persistent Endpoint
- Per-second billing
- Deployment logs & status monitoring
- Custom images
- Priority support
Business
$199/month
For teams with high-throughput GPU workloads
$50 credits/month included
Start Business Trial- Everything in Pro
- Up to 30 concurrent GPUs
- Up to 5 Persistent Endpoints
- Usage analytics dashboard
- Dedicated support
Enterprise
Custom
For organizations with advanced needs
Custom credit packages available
Contact Sales- Everything in Business
- Unlimited concurrent GPUs
- Volume discounts on Endpoints
- Dedicated support engineer
- SLA guarantees
- Custom contracts