vast-ai-python

name	vast-ai-python
description	Rent GPU instances on Vast.ai marketplace using Python SDK. Search offers, create on-demand/interruptible instances, manage lifecycle, templates, volumes, data transfer, cloud sync. Use when mentioning vast.ai, VastAI, GPU rental, spot GPU, interruptible GPU instances, marketplace GPU.

Vast.ai Python SDK

GPU marketplace — rent from individual hosts and datacenters. Python SDK wraps the CLI as typed methods via from vastai import VastAI.

Install & auth

pip install vastai

from vastai import VastAI
vast = VastAI(api_key="YOUR_KEY")  # or omit to read ~/.config/vastai/vast_api_key

Constructor params: api_key, server_url, retry=3, raw=False, quiet=False.

Core workflow

1. search_offers(query=..., order=..., type=..., limit=...) → find GPU
2. create_instance(id=OFFER_ID, image=..., disk=..., ssh=True, direct=True) → rent it
3. Poll show_instance(id=...) until actual_status == "running"
4. ssh_url(id=...) or copy(src=..., dst=...) → connect / transfer data
5. stop_instance(id=...) (pause, storage charges continue) or destroy_instance(id=...) (delete all)

When to load references

• Choosing instance type (on-demand vs interruptible vs reserved): see instance-types.md
• Searching & filtering offers, query syntax, DLPerf/reliability scores: see searching-offers.md
• Creating instances via SDK or REST API, runtype options: see creating-instances.md
• Instance lifecycle — start/stop/destroy, status states, billing: see managing-instances.md
• Templates — creating, Docker images, launch modes, on-start scripts: see templates.md
• Storage — container disk, volumes, data movement, cloud sync: see storage-and-data.md
• Pricing model, cost components, billing: see pricing.md