Browsing GPUs
Step 1: Choose a GPU type
The marketplace starts with a card grid showing available GPU families. Each card displays:- GPU model name (e.g., RTX 4090, H100)
- VRAM
- Architecture (Hopper, Ampere, Ada Lovelace, etc.)
- Cheapest available price
- Number of available offers
- In-stock indicator
Step 2: Filter and compare offers
Once you’ve selected a GPU type (or skipped to browse all), the offers table shows every matching offer with these columns:
The table is sorted by price (lowest first) by default. Click column headers to sort by VRAM or price.
Filters
Narrow results with these filters:
Active filters are shown with a count badge. Click Clear All Filters to reset.
Pricing
All offers show an effective hourly rate that includes everything. The 10% platform fee is already included in the displayed price. What you see is what you pay.Price components
When you click Deploy on an offer, the deploy page shows a full breakdown:- GPU rate — Base hourly rate for the GPU hardware
- Disk cost — Storage cost per hour
- Public IP — IP allocation cost (if applicable)
- Platform fee — 10% platform fee
Bandwidth costs
Some offers may have additional usage-based costs for data transfer (egress and ingress). These are noted during the deploy flow where applicable and are not included in the hourly rate.Offer details
Clicking Deploy on an offer takes you to the deploy configuration page where you’ll see a full cost breakdown before committing. The marketplace only shows offers that are:- Currently available and in stock
- On-demand pricing (no spot/preemptible)