Vast.ai
GPU marketplace where independent hosts and data centers rent out GPUs, with on-demand, interruptible and reserved instances, serverless endpoints and clusters controlled from a console, CLI, Python SDK or REST API.
About Vast.ai
Training, fine-tuning and running open models needs GPUs, and fixed-price clouds can be expensive or short on the exact GPU type you want.
Who it’s for
- AI developers and researchers who want cheap hourly GPUs for experiments, fine-tuning or batch jobs
- Teams and agents that provision GPU compute programmatically and can tolerate varying hosts
When to consider it
Consider Vast.ai when you want to compare live offers across 68+ GPU types, from an RTX 3060 to a B200, and pay per second, using interruptible instances for fault-tolerant work or reserved terms for steady workloads.
Tradeoffs & limitations
- There is no fixed price list: hosts set their own prices and rates change with supply and demand. You add credit before starting (from $5); interruptible instances are described as often 50%+ cheaper but can be paused, and reserved 1, 3 or 6 month terms give up to 50% off.
- GPU time is billed per second, but storage keeps being charged while a stopped instance exists and bandwidth is charged per byte at each host's rate; when credit runs out instances are stopped, and without a saved card they are scheduled for deletion.
SOTA overview · Documentation-based assessment · Sources & review method
Updates
No updates shared yet.
Discussion
Newest firstAsk a question or share how you use Vast.ai.
Keep it helpful. Community rules
Loading discussion…
Sources & review method
Documentation-based assessment · Nov 2, 2026 · Prepared with AI assistance; not a hands-on benchmark.
Checked by SOTA · AI-assisted documentation review. Selection advice is our assessment; verify current requirements for your deployment.
Import history & original evidence
Vast.ai official website
GPU marketplace where independent hosts and data centers rent out GPUs, with on-demand, interruptible and reserved instances, serverless endpoints and clusters controlled from a console, CLI, Python SDK or REST API.
Based on official pages and announcements checked on 2026-11-02. No hands-on test, performance benchmark or popularity ranking is claimed.