# Ona (formerly Gitpod)

Official pricing: https://ona.com/pricing  
Category: dev-env · Isolation: vm

## Pricing regimes (raw)

- **Ona Cloud (Core), Standard 4 vCPU / 16 GB = 1 OCU/h** (sizes): Standard (4 vCPU / 16 GB, 1 OCU/h) 4 vCPU/16 GiB $0.25/h
- **Ona Cloud (Core) GPU VM, 16 vCPU / 64 GB = 7 OCU/h** (sizes): GPU / Data Science VM (16 vCPU / 64 GB / 300 GB, 7 OCU/h) 16 vCPU/64 GiB $1.75/h
- **Enterprise runner in own AWS VPC (customer pays EC2 on-demand)** (sizes): Small (m6i.large) 2 vCPU/8 GiB $0.096/h; Regular (m6i.xlarge) 4 vCPU/16 GiB $0.192/h; Large (m6i.2xlarge) 8 vCPU/32 GiB $0.384/h; Extra Large (m6i.8xlarge) 32 vCPU/128 GiB $1.536/h [sales, alt]
- **Enterprise runner in own AWS VPC, GPU Large (g5.4xlarge, 1× A10G)** (sizes): GPU Large (g5.4xlarge, 1× A10G 24 GB) 16 vCPU/64 GiB $1.624/h [sales, alt]

## Features

Yes: snapshots, pause/resume, persistent disk, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, full VM (own kernel), Docker inside, root, browser, preinstalled agents, SSH, HTTPS preview URLs, open internet, BYOC, SOC 2, SSO, GPU, webhooks, scheduled runs (cron), EU data residency, spend limits, agent harness API, gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables)

No: memory snapshots, fork/clone, volumes, nested virtualization, desktop GUI, computer-use API, code interpreter, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, public IPv4, custom domains, raw TCP inbound, egress allowlist, static egress IP, private networking, self-hosting, open source, HIPAA, arm64, Windows, macOS, Python, Node.js, GPU desktop, HTTP method/path egress rules, wake on request, live resize, live fork (no pause), memory fork, audit logs, MCP server, automatic snapshots, snapshots on demand, hosted agent API (their own agent), secret proxy, secret proxy for any API, extra volumes, shared volumes

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

## Caveats

- OCU rates are published for only two classes: Standard 4 vCPU/16 GB = 1 OCU/h and GPU 16 vCPU/64 GB = 7 OCU/h. Core supports up to 32 vCPU/128 GB/200 GB, but the OCU rates for other classes are unpublished. A 2025 Ona story quoted 2 vCPU/8 GB at $0.11/h and 4/16 at $0.23/h under earlier pricing (superseded).
- OCU price: $0.25 on the $20-$200 tiers and for top-ups (from $10 / 40 OCU); $0.2273 on the $500/2,200-OCU tier. Monthly OCUs expire; top-ups last 1 year but expire if the subscription ends.
- OCUs also pay for Ona agent LLM tokens, from about 1 OCU (explain a small codebase) to 8 OCU (add a feature). A connected Codex/ChatGPT plan is not charged OCUs, and Claude Code usage is billed by Anthropic.
- Billing granularity and per-start minimum for Ona Cloud are unpublished (null).
- Stopped/archived storage: the pricing FAQ says inactive environments 'generate storage costs based on VM disk size', but the usage docs only list OCU consumption while running. No storage OCU rate is published (null). Core archives stopped environments after 3 days and deletes them 4 days later (7 days in total), so long-lived stopped state is impossible on Core.
- Per-class disk size on Ona Cloud is unpublished (included_disk_gib null). The Core maximum is 200 GB, and 300 GB for the GPU VM.
- Egress pricing is unpublished (null). There is no public IPv4 per environment; ports are exposed via TLS URLs.
- Programmatic API/SDK access is Enterprise-only per the pricing comparison, which limits Core as an API sandbox backend.
- The Enterprise fee ($60,000/12 months) comes from vendor listings on AWS Marketplace, not ona.com. It is for one 'unit' whose definition is unpublished, and private offers vary. Enterprise-mode hourly prices are AWS EC2 list prices (via the third-party instances.vantage.sh mirror) paid to AWS, not to Ona. EBS, snapshot and GCP costs are not encoded.
- Failed payment: 7-day grace, then the account is locked and remaining Core credits are forfeited.
- No perf benchmark data for Ona in this research set.
- Ona is now part of OpenAI (site banner). Watch for pricing changes.
- Mode-plan coupling (verifier 2026-09-28): Ona Cloud modes are sold only on Core plans, self-hosted AWS modes only on Enterprise (mode.plans). Without this, with sales+alt opted in, the engine could price customer-paid EC2 under a $20 Core plan and skip the Enterprise contract fee.
- Per-tier OCU amounts ($50/200, $100/400, $200/800, $500/2,200) are rendered client-side on ona.com/pricing and were not re-verifiable from static HTML on 2026-09-28; the page confirms '80 - 2,200 OCUs (monthly recurring)', 'from $10 / 40 OCUs' and Core 'from $20/month'.
- Auto-stop (fix 2026-09-28): new environments auto-stop after 30 min without activity by default; only SSH sessions, IDE use, running agents and CLI keep-alives count, background processes (builds, servers, cron) do NOT. Org policy can set the max to 30 min/1 h/3 h/8 h/none; auto-stop can be disabled ("Never"/0s). An always-on devloop box needs auto-stop disabled or it stops.
- Boot time of an Ona Cloud environment is not published (boot_overhead_s left null); category stays dev-env (a cloud dev environment, not a sandbox API; Core has no API/SDK access).
- Prebuild snapshots on Ona Cloud are free for now ('this may change'); prebuild environment runtime is billed as normal OCU usage; prebuilds are deleted after 5 days. On Enterprise the customer pays EBS snapshot storage for actual data stored; Ona's warm-pool doc estimates gp3 EBS at roughly $0.08/GB-month and confirms EC2 class costs (~$70 / $140 / $280 per running month for Small/Regular/Large), matching the card's $0.096/$0.192/$0.384 per hour.
- OpenAI acquisition closed on 2026-08-10 (Ona is now part of OpenAI, Codex team). No pricing change announced as of 2026-09-29.
- Ona also meters Enterprise environment runtime in its own credits (usage type USAGE_TYPE_ENVIRONMENT alongside USAGE_TYPE_AGENTIC), and base-snapshot builds are metered as normal environment usage (warm pools consume no credits). The credit rate per environment-hour is unpublished, so the enterprise modes (EC2 price only) may understate the Ona-side cost.
- Per-user environment caps exist as an org policy: 'Maximum concurrent' running environments per user and 'Maximum total' (running or stopped), defaults 10 concurrent / 50 total, configurable 1-100. The policy page does not say whether the defaults are enforced on Core (the policies overview says most policies require Enterprise). Core is marketed as 'Unlimited parallel environments' org-wide, but a single-user/service-account agent workload may hit 10 concurrent.

## How this provider charges


Ona (now part of OpenAI) sells cloud dev environments plus background agents. It has two regimes. **Core**
runs on Ona Cloud (multi-tenant, AWS eu-central-1 / us-east-1) and is billed in prepaid **Ona Compute Units
(OCUs)**. The same OCUs pay for environment runtime and for Ona-agent LLM tokens. **Enterprise** is
contract-priced and runs Ona-managed runners inside the customer's own AWS/GCP VPC, so the customer pays the
cloud bill for the VMs directly.

OCU price: **$0.25/OCU** on the $20–$200 tiers and for top-ups, **$0.2273/OCU** on the $500 tier.
Standard env 4 vCPU / 16 GB = **1 OCU/h = $0.25/h** ($0.227/h at the $500 tier). GPU VM 16 vCPU / 64 GB = **7 OCU/h = $1.75/h**.

## Regime table

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| Core $20 / $50 / $100 / $200 | Default self-serve (new orgs start on Core) | Monthly subscription converted to OCUs. Monthly OCUs expire at the end of the billing month (no rollover). The variant can't be changed without cancel + resubscribe | $20/80, $50/200, $100/400, $200/800 OCU ($0.25/OCU). Up to 100 members, no per-seat fee. Unlimited parallel envs. 50 automations / 10 parallel actions | https://ona.com/pricing, https://ona.com/docs/ona/billing/overview |
| Core $500 | Heavy teams | Same, with 10% bonus OCUs | $500 / 2,200 OCU = $0.2273/OCU | https://ona.com/pricing |
| Top-up (manual) | Monthly OCUs exhausted | One-time prepaid purchase. Top-ups are used after subscription OCUs, valid up to 1 year, and expire if the subscription ends | from $10 / 40 OCU ($0.25/OCU) | https://ona.com/pricing, https://ona.com/docs/ona/billing/usage |
| Auto top-up | Core, opt-in | Card charged when the balance drops below 20 OCU. The amount is chosen by the user, with a cooldown between charges | $0.25/OCU (amount options not published) | https://ona.com/docs/ona/billing/overview |
| Bonus / gift / coupon credits | Promotions, coupon at signup | Consumed last, usually a longer expiry | Amounts not published. A third party reports "$100 extra credit for signing up in September [2025]" | https://ona.com/docs/ona/billing/usage, third-party: thefridaydeploy.substack.com |
| Standard environment (Ona Cloud) | Running env, 4 vCPU / 16 GB | OCUs per running hour by class. Granularity/minimum not published | 1 OCU/h = $0.25/h | https://ona.com/docs/ona/billing/usage |
| GPU-accelerated environment (Ona Cloud) | Running GPU VM, 16 vCPU / 64 GB / 300 GB | Same | 7 OCU/h = $1.75/h. GPU model not stated (the AWS-runner equivalent g5.4xlarge = 1× A10G) | https://ona.com/docs/ona/billing/usage, https://ona.com/pricing |
| Other Core classes (up to 32 vCPU / 128 GB / 200 GB) | Larger/smaller envs | OCUs per hour | **Not published** (null) | https://ona.com/pricing |
| Agent usage (Ona agent) | Prompting Ona agents / automations | OCUs from the same pool | ~1 OCU (explain small codebase), 3 (large), 4 (new web app), 5 (bug fix, large codebase), 8 (feature, medium codebase) | https://ona.com/pricing, https://ona.com/docs/ona/billing/usage |
| Bring your own agent | Connected Codex/ChatGPT plan, or Claude Code installed in the env | No OCUs for model usage. The env still burns OCUs while running | $0 Ona model charge | https://ona.com/docs/ona/billing/usage, https://ona.com/pricing |
| Stopped env | After auto-stop / manual stop | Pricing FAQ: inactive envs "generate storage costs based on VM disk size". No storage OCU rate is published. Core archives after 3 days and deletes 4 days later | null | https://ona.com/pricing, https://ona.com/docs/ona/environments/archive-auto-delete |
| Auto-stop | All envs | Env keeps billing until the timeout fires | Default 30 min. Options 30 min / 1 h / 3 h / 8 h / Never. Orgs can cap the max | https://ona.com/docs/ona/organizations/policies/environment-timeout |
| Enterprise (contract) | Security/VPC, SSO, API/SDK, warm pools | Annual contract with "custom credits" (Enterprise uses credits, not OCUs) | AWS Marketplace: 12 months, **$60,000 TCV** per unit (unit undefined), non-refundable. Private offers available. ona.com says "Custom pricing" | https://aws.amazon.com/marketplace/pp/prodview-752jqvg74yo7k, https://aws.amazon.com/marketplace/pp/prodview-vn4wpn7u4afrq |
| Enterprise self-hosted runner compute (AWS) | Enterprise only | Customer's AWS bill: EC2 per environment + EBS + base-snapshot builds. Runner baseline < $8/month | Default classes: Small m6i.large 2/8/45 GB $0.096/h, Regular m6i.xlarge 4/16/80 GB $0.192/h, Large m6i.2xlarge 8/32/100 GB $0.384/h, Extra Large m6i.8xlarge 32/128/200 GB $1.536/h, GPU Large g5.4xlarge 16/64/300 GB $1.624/h (AWS us-east-1 list, **third-party mirror** instances.vantage.sh) | https://ona.com/docs/ona/runners/aws/environment-classes, https://ona.com/docs/ona/runners/aws/aws-runner-costs |
| Enterprise spot classes | Enterprise AWS runner | AWS spot price (variable). On reclamation the env is re-provisioned and EBS reattached (data kept) | Extra Large Spot m7i.8xlarge, GPU Large Spot g5.4xlarge. Price varies | https://ona.com/docs/ona/runners/aws/environment-classes |
| Enterprise GCP runner | Enterprise | Customer's GCP bill | not encoded | https://ona.com/docs/ona/runners/ona-cloud |
| Free (legacy) | 2025 only | No fee | Up to 3 parallel envs, ≤ 4 vCPU/16 GB (Ona migration story). "$10 free credit" is third-party only. **Not on the current pricing page** | https://ona.com/stories/migrating-from-classic-to-ona |
| Failed payment | Card fails | 7-day grace, then the account is locked and remaining Core credits are forfeited | — | https://ona.com/docs/ona/billing/overview |

Dated changes: Gitpod Classic sunset 2025-10-15 (Classic PAYG was $0.36/h for 4 cores/8 GB). Ona's 2025 launch pricing
quoted 2 vCPU/8 GB $0.11/h and 4/16 $0.23/h (superseded by OCUs). Ona joined OpenAI (2026).

## Gotchas

1. **One pool for compute and AI.** Agent conversations draw from the same OCUs as env-hours. A single "add a feature"
   prompt (~8 OCU) costs as much as 8 hours of a Standard env. Using your own agent (Claude Code, or a connected Codex plan) avoids this.
2. **Only two classes have published rates.** Core advertises up to 32 vCPU / 128 GB, but only 4/16 (1 OCU/h) and the 16/64
   GPU VM (7 OCU/h) have public OCU rates. Budgeting for other sizes is guesswork.
3. **No 4/8 shape.** The smallest priced class is 4 vCPU / **16 GB**, so a 4/8 workload pays for 16 GB.
4. **Prepaid, expiring credits.** Monthly OCUs vanish at month end. Buying a bigger tier "to be safe" wastes money, and the
   tier can't be changed without cancelling. Top-ups die with the subscription.
5. **Idle tail is billed.** Auto-stop defaults to 30 min but can be set to Never. Background processes (builds, servers) don't keep an
   env alive unless you use the CLI keep-alive. An open SSH/IDE session or a running agent session does keep it alive and billing.
6. **Aggressive deletion on Core.** Stopped envs are archived after 3 days and deleted 4 days later (7 days total). Anything
   uncommitted is lost, so long-lived paused state is not possible on Core.
7. **API/SDK is Enterprise-only.** Core is a human/agent IDE product, not a programmable sandbox API.
8. **Enterprise means two bills.** You pay Ona's contract (the Marketplace listing shows a $60k/yr unit) plus your own AWS/GCP compute
   and EBS. Compute is **not** included.
9. **Granularity unknown.** Ona doesn't publish per-second vs per-minute OCU metering or a minimum.

## Worked example

Workload: 4 vCPU / 8 GiB, 50 concurrent × 8 h/day × 22 days = **8,800 env-hours**, 30% CPU, 50 GiB snapshots,
100 GiB egress. The smallest priced class is Standard 4 vCPU/16 GB = 1 OCU/h, so it needs **8,800 OCU**. CPU utilisation doesn't matter (billed while running).
Snapshots: no published storage rate, and Core deletes stopped envs after 7 days, so this is treated as $0 but unknown. Egress: unpublished (unknown).
No Ona-agent tokens are assumed (own agent).

| Regime | Feasible? | Monthly total |
|---|---|---|
| Core $20 + top-ups | Yes (unlimited parallel envs) | $20 + 8,720 × $0.25 = **$2,200** |
| Core $50–$200 + top-ups | Yes | **$2,200** (all at $0.25/OCU) |
| Core $500 + top-ups | Yes | $500 + 6,600 × $0.25 = **$2,150** (cheapest) |
| Core $500, default 30-min auto-stop tail each session | Yes | +550 OCU → **$2,287.50** |
| Core $500, plus Ona-agent use equal to runtime (×2 OCU) | Yes | $500 + 15,400 × $0.25 = **$4,350** |
| Anti-pattern: timeout "Never", envs left running 24 h × 22 d | Yes | 26,400 OCU → $500 + 24,200 × $0.25 = **$6,550** |
| Enterprise self-hosted AWS (Regular m6i.xlarge 4/16) | Contract | $5,000 (Marketplace $60k/yr ÷ 12) + 8,800 × $0.192 = $1,689.60 AWS EC2 → **≈ $6,690** + EBS (50 × 80 GB) + snapshots + egress at AWS rates |
| Free (legacy) | No: max 3 parallel envs, and discontinued | n/a |