Paperspace Pricing Calculator 2026

$8/month subscription plus storage

Individual ML practitioner using Gradient Pro for consistent GPU access to A4000-class (16–24GB VRAM) machines for model fine-tuning, stable diffusion, and experimentation. Requires additional storage spend for model checkpoints.

$39/month flat rate

Heavy AI/ML user needing daily access to high-VRAM machines (up to 80GB VRAM, including A100-class) for larger model training runs. 50GB persistent storage included. Sessions capped at 6 hours each.

$123.60 (40 hours × $3.09/hr A100)

Team running a single large fine-tuning job on an A100 GPU for 40 hours without a subscription. Appropriate for infrequent compute needs where a monthly plan would not break even.

$119.00 (20 hours × $5.95/hr H100)

Research team running a compute-intensive training job on an H100 GPU for 20 hours. H100 availability at Paperspace confirmed at $5.95/hr in third-party GPU pricing comparisons.

~$8–$10/month (10 hrs × $0.82/hr = $8.20 compute + minor storage)

A solo researcher running occasional experiments on an RTX5000 GPU for approximately 10 hours per month, with minimal persistent storage.

$65.20–$75.20/month total ($55.20 compute + $10–$20 storage)

A creative professional running Stable Diffusion or similar workloads on an A5000 GPU for approximately 40 hours per month. Includes storage for model files.

~$39–$49/month flat (plan fee + any overage storage beyond included amount)

A user on Paperspace's Growth subscription plan who uses GPUs up to 50GB VRAM daily for image generation or model training, leveraging unlimited usage with 50GB persistent storage.

~$119/run (20 hrs × $5.95/hr)

A team running a single intensive model training job on Paperspace's H100 GPU for 20 hours.

$50 for approximately 24 hours of unattended runtime (user-reported)

A new user sets up a machine, forgets to shut it down, and is billed for 24+ hours of continuous runtime with no active usage.

$60/month for 1TB of persistent storage (user-reported)

User signs up on the free tier, attaches a large disk volume, uses the machine briefly, then forgets about the storage — which continues billing monthly.

£79 for a single failed 30-minute deployment (user-reported)

User attempts a model deployment that fails to build after 30 minutes; compute time is billed regardless of whether the deployment succeeded or the service became usable.

$980/month (effective ~$1.30/hr vs $2.30/hr on-demand)

Dedicated V100 GPU at a fixed monthly rate for teams with sustained workloads running 20+ hours per day, avoiding on-demand rate volatility.

$119/month (20 hours × $5.95/hr)

Ad hoc high-end training using the H100 at $5.95/hr for short bursts, e.g. 20 hours/month for experimentation.

$18/month (40 hours × $0.45/hr)

Light usage of the M4000 at $0.45/hr for occasional inference or smaller fine-tuning tasks, e.g. 40 hours/month.