VS Code in browser, Jupyter Lab, Jupyter + PyTorch, or a full dev environment with persistence — ready in 2–5 minutes, no SSH required.
One-click GPU dev environments on packet.ai launch VS Code in your browser, Jupyter Lab, or a full dev environment with persistence — ready in 2–5 minutes from the same dashboard you use to provision GPUs.
Key takeaways
GPU time is expensive. The previous workflow — provision, SSH in, install tools, configure ports, start services — cost 20–30 minutes of setup before any actual work. Multiplied across a team, that’s real money. One-click environments eliminate all of it.
VS Code in Browser
Full VS Code via code-server, accessible from any browser. Extensions, terminal, Git — everything VS Code has, no local install required.
Jupyter Lab
Jupyter Lab with numpy, pandas, and matplotlib pre-loaded. Ready for exploratory work and data analysis immediately on boot.
Jupyter + PyTorch
Jupyter with PyTorch and CUDA pre-configured. Start training or fine-tuning immediately — no driver setup, no CUDA configuration.
Persistent Workspace
Home directory survives pod restarts. Detach the GPU to stop paying, reattach and continue exactly where you left off.
Full Dev Environment
VS Code + Jupyter + Persistent Workspace combined. The recommended default for any development workflow longer than one session.
When you select an environment and launch your GPU, a startup script installs and configures the selected services. Each service is registered as a systemd unit, so it starts automatically on pod boot and restarts on failure — no manual service management.
When setup completes, packet.ai detects which ports the services are listening on and creates authenticated proxy URLs. The URLs appear in your dashboard automatically. No port forwarding, no SSH tunnel configuration.
Time to productive
Select environment → launch GPU → wait 2–5 minutes for setup → click the URL in dashboard → writing code. The entire setup that used to take 20–30 minutes now happens in the background while the GPU boots.
At packet.ai’s GPU rates ($0.66/hr for RTX PRO 6000, $2.25/hr for H200, $3.75/hr for B200), every 30 minutes of setup costs $0.33–$1.88 before any actual work happens. For a team of 5 engineers each doing this once per day, five days per week: that’s $1.65–$9.40/day, or $430–$2,444/year in setup GPU time. One-click environments eliminate that cost entirely.
The bigger cost is context-switching time. 20 minutes of configuration means the mental model of the problem you were solving has gone cold. Productive sessions start immediately, not after a setup ritual.
Last reviewed: 10 June 2026. Browse GPU clusters on packet.ai →
Same models. Same API. Fraction of the cost. Start free — no credit card required.
Start Building →