# Paperspace

Official pricing: https://docs.digitalocean.com/products/paperspace/pricing/  
Category: gpu-cloud · Isolation: vm

## Pricing regimes (raw)

- **Current Core CPU hourly** (sizes): C1 1 vCPU/0.5 GiB $0.0045/h; C3 2 vCPU/2 GiB $0.018/h; C4 2 vCPU/4 GiB $0.04/h; C5 4 vCPU/8 GiB $0.08/h; C6 8 vCPU/16 GiB $0.16/h; C7 12 vCPU/30 GiB $0.3/h
- **RTX-A4000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **RTX-A5000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **RTX-A6000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **V100 16GB, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **P6000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **RTX5000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h
- **P5000, on-demand (×1)** (resource): $0/vCPU-h, $0/GiB-h

## Features

Yes: snapshots, pause/resume, persistent disk, volumes, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), Docker inside, SSH, public IPv4, open internet, static egress IP, private networking, SOC 2, HIPAA, SSO, GPU, Python, EU data residency, spend limits, snapshots on demand, gVisor or VM (no shared kernel), firewall inside (nftables), extra volumes, shared volumes

No: memory snapshots, fork/clone, your own Docker/OCI image, browser, desktop GUI, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, preinstalled agents, HTTPS preview URLs, egress allowlist, self-hosting, BYOC, open source, arm64, Windows, macOS, Node.js, HTTP method/path egress rules, wake on request, live resize, live fork (no pause), memory fork, MCP server, automatic snapshots, agent harness API, hosted agent API (their own agent), inbound access rules, secret proxy, secret proxy for any API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/paperspace.json → feature_evidence

## Caveats

- Public DigitalOcean-acquired VM/GPU platform; Core hourly/monthly, Gradient notebook/deployment plans. Marketing page contains conflicting Pro $8 and $12/mo in separate sections. Keep product-specific compute tables in evidence; do not combine a notebook fee with unrelated Core tariff.
- Current DigitalOcean docs supersede marketing tables: C5 $0.08/hour, 50GB disk separately $5/mo cap; not free included disk. All compute hourly only while powered on; storage/IP keep billing when off.
- Windows templates unavailable to new users since 2024-07-01, so card windows=false for new buyers; legacy users may retain Windows access. Marketing Standard $35 Windows offer must not be ranked as current public availability.
- Pro $8 personal vs $12/team user is resolved by current docs; Gradient subscriptions are not required Core subscription. Storage/snapshot prices are capacity tiers with hourly accrual/monthly caps, not linear GiB rates; left null in engine and explicit example.
- Current multi-GPU table and prose conflict for V100×8 and A6000×8; verify actual SKU before extrapolating single GPU rates.
- Null values mean unknown, not free/unlimited. No model-token cost, tax or unverified ancillary charge included.
- Native browser/device/task/credit prices with unknown CPU/RAM are deliberately non-priceable. Check regime before interpreting any partial resource estimate.
- GPU modes grafted from the v4-gpu research card (per-GPU rates; exact rows in research/gpu.json).
- Boot disk (verified 2026-09-28): block storage is tiered with hourly accrual and a monthly cap: 50 GB $5/mo, 100 GB $7, 250 GB $10, 500 GB $25, 1 TB $55, 2 TB $120; billed while the machine exists, even when off. Card encodes the 50 GB tier rate ($0.10/GiB-month) linearly: it overstates larger disks (250 GB is $0.04/GiB) and understates disks under 50 GB (a 50 GB tier is the minimum). Snapshots are $1-$8/mo per tier (not modelled).
- Desktop/browser (verified 2026-09-28): Linux machines are reached by SSH only; desktop streaming is 'only available for your Windows-based machines' and Windows templates are closed to new users. No managed browser product. desktop = false, browser = false.
- Boot time is not published (boot_overhead_s left null). Paperspace machines are plain VMs with no sandbox API; the engine's 'plain VM' compromise applies for sandbox-style presets.
- Older GPUs (M4000, P4000/P5000/P6000, RTX4000/5000, V100) are still listed and priced on the official pricing and machine-types pages (2026-09-28), with no deprecation notice, so they are NOT flagged legacy. V100 is listed as V100/V100-32G $2.30/hr and is available on the Growth plan only per the pricing table.
- Linux (terminal/SSH-only) machines REQUIRE a public IP: dynamic IP adds $0.0045/h while running (static $0.0045/h, $3/mo cap, billed until deleted). Effective hourly = size rate + $0.0045 (C1 $0.0045 -> $0.009, +100%; C5 $0.08 -> $0.0845, +5.6%). Plus the mandatory >=50 GB boot disk at $0.0074/h (cap $5/mo) for as long as the machine exists.
- Public bandwidth: sustained >=1 Gbps for >10 s is rate-limited to 200 Mbps unless a ticket removes it; public throughput capped at 2 Gbps in NY2, 900 Mbps elsewhere. No IPv6.
- Multi-GPU table labels 'V100-32G x8 $9.20' and 'A6000 x8 $7.56' are 4x the single rate and the machine-types page lists only x2/x4 for these; treat as x4. A100-80G x2/x4 discontinued 2024-03-20 (x8 only). A100 x8 $24.72, A100-80G x8 $25.44, H100 x8 $47.60 (8 GPUs only).

## How this provider charges


YC: Winter 2015; directory status **Acquired**. Research scope: public-product-researched. Source: https://www.ycombinator.com/companies/paperspace

Paperspace Core and Gradient

## Regime table

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| Current Core CPU hourly | Public listed mode | sizes; CPU alloc, memory alloc | C1: 1 vCPU/0.5 GiB $0.0045/h; C3: 2 vCPU/2 GiB $0.018/h; C4: 2 vCPU/4 GiB $0.04/h; C5: 4 vCPU/8 GiB $0.08/h; C6: 8 vCPU/16 GiB $0.16/h; C7: 12 vCPU/30 GiB $0.3/h; C8: 16 vCPU/60 GiB $0.6/h; C9: 24 vCPU/120 GiB $0.9/h; C10: 32 vCPU/244 GiB $1.6/h | https://docs.digitalocean.com/products/paperspace/pricing/ |
| Core usage | Plan limits apply | Monthly fee; credits only as stated | fee=0; included usage=$0; concurrent=None; max session h=None | https://docs.digitalocean.com/products/paperspace/pricing/ |
| Retained disk / GiB-month | Lifecycle/scope must be checked | Native add-on | unverified, not $0 | https://docs.digitalocean.com/products/paperspace/pricing/ |
| Snapshots / GiB-month | Lifecycle/scope must be checked | Native add-on | unverified, not $0 | https://docs.digitalocean.com/products/paperspace/pricing/ |
| Egress / GiB | Lifecycle/scope must be checked | Native add-on | $0 | https://docs.digitalocean.com/products/paperspace/pricing/ |
| IPv4 / month | Lifecycle/scope must be checked | Native add-on | $3 | https://docs.digitalocean.com/products/paperspace/pricing/ |

## Verified billing facts and gotchas

1. Public DigitalOcean-acquired VM/GPU platform; Core hourly/monthly, Gradient notebook/deployment plans. Marketing page contains conflicting Pro $8 and $12/mo in separate sections. Keep product-specific compute tables in evidence; do not combine a notebook fee with unrelated Core tariff.
2. Current DigitalOcean docs supersede marketing tables: C5 $0.08/hour, 50GB disk separately $5/mo cap; not free included disk. All compute hourly only while powered on; storage/IP keep billing when off.
3. Windows templates unavailable to new users since 2024-07-01, so card windows=false for new buyers; legacy users may retain Windows access. Marketing Standard $35 Windows offer must not be ranked as current public availability.
4. Pro $8 personal vs $12/team user is resolved by current docs; Gradient subscriptions are not required Core subscription. Storage/snapshot prices are capacity tiers with hourly accrual/monthly caps, not linear GiB rates; left null in engine and explicit example.
5. Current multi-GPU table and prose conflict for V100×8 and A6000×8; verify actual SKU before extrapolating single GPU rates.

Native pricing (not coerced into incompatible engine units):
```json
{
  "block_50gb_hour": 0.0074,
  "block_50gb_month_cap": 5,
  "snapshot_50gb_hour": 0.0015,
  "snapshot_50gb_month_cap": 1,
  "public_ip_hour": 0.0045,
  "public_ip_month_cap": 3,
  "gradient_personal_pro_month_cap": 8,
  "gradient_team_pro_month_cap": 12
}
```

## Worked example

Requested: 4 vCPU/8 GiB, 50 simultaneous × 8 h/day × 22 days = **8,800 instance-hours**, 30% CPU; 50 GiB snapshots and 100 GiB egress. A month is 730 hours only when explicitly used in rate conversion.

Current C5 4/8 hourly: 8,800×$0.08=$704. Fifty minimum 50GB block disks retained all month add 50×$5=$250; one 50GB snapshot adds $1; 100GB egress $0. Subtotal $955, or $1,105 with fifty retained public IPs ($3 each). Fifty machines require account quota approval; price is not availability guarantee.

## Sources and coverage

- https://www.ycombinator.com/companies/paperspace
- https://docs.digitalocean.com/products/paperspace/pricing/
- https://docs.digitalocean.com/products/paperspace/machines/details/machine-types/
- https://docs.digitalocean.com/products/paperspace/machines/details/limits/
- https://www.paperspace.com/pricing
- https://www.paperspace.com/

Page captures and documentation indexes/bundles are in `raw/`; provider JSON lists the evidence paths. HN query results were captured separately as `<id>--hn.txt`; they are discovery leads, not current tariff authority. Downloaded docs do not imply every feature was verified; unsupported assertions remain null.