Home / Compare / OpenAI Codex cloud vs Amp
OpenAI Codex cloud vs Amp
Verdict (2026-11-15): OpenAI Codex cloud scores higher: 69/100 vs 53/100 as coding agents with cloud machines. Choose OpenAI Codex cloud for chatGPT subscribers who want cloud coding tasks without paying for machines separately. Choose Amp for running many agent tasks in parallel, each in its own paused-when-idle cloud machine.
OpenAI Codex cloud
Highest scoreOpenAI's coding agent running tasks in cloud VMs, included in ChatGPT Plus, Pro, Business and Enterprise plans.
Amp
Coding agent that gives every thread its own cloud machine (an orb), with published per-size hourly prices.
Score by part
Cost at four workloads
| Workload | OpenAI Codex cloud | Amp |
|---|---|---|
| Short tasks: 1 vCPU, 1 GiB, 50 h/month | — | $3.33 |
| Dev agent: 2 vCPU, 4 GiB, 176 h/month | — | $20.00 |
| Fleet: 2 vCPU, 4 GiB, 1,760 h/month | — | $187.26 |
| Always on: 4 vCPU, 8 GiB, 730 h/month | — | $137.58 |
Per month, cheapest plan for each, without storage, network or GPU. Lowest cost in bold.
Choose OpenAI Codex cloud if
- ChatGPT subscribers who want cloud coding tasks without paying for machines separately.
- Teams that need per-environment network allowlists, VPN access or OIDC to cloud accounts.
- Starting and following tasks from phone, web, Slack or Linear.
Choose Amp if
- Running many agent tasks in parallel, each in its own paused-when-idle cloud machine.
- Developers who want to use their own ChatGPT or Claude subscription or API keys for models.
- Teams that want to run the same agent on their own machines at no extra cost.
Side by side
| OpenAI Codex cloud | Amp | |
|---|---|---|
| Category | coding agent | coding agent |
| Our score | 69/100 | 53/100 |
| Isolation | Each cloud task runs in its own VM | Separate sandboxed cloud machine per thread; Amp's security page says e2b provides the compute instances. The isolation technology is not stated by Amp. |
| vCPU per hour | — | $0.0504 |
| Memory per GiB-hour | — | $0.0162 |
| Billing | No separate machine billing; cloud tasks use the plan's usage allowance | per minute |
| Free tier | No free cloud tasks. The Free ($0) and Go ($8/month) ChatGPT plans list Codex only in the desktop app; cloud tasks start with Plus. | No free hosted orb time. The Hobby tier itself is free, and self-hosted orbs or your own runners cost nothing on Hobby, Individual and Teams tiers, but Amp-hosted orbs are pay-as-you-go. |
| Pause and resume | Yes | Yes |
| Snapshots | Yes | Yes |
| Network controls | Yes | Not stated |
| Max vCPU | 4 vCPU | 16 vCPU |
| Max memory | 16 GiB | 44 GiB |
| Open source | No | No |
| Self-host | No | Yes |
| SDKs and interfaces | Web app, Desktop and mobile apps, CLI, Codex SDK | CLI, Web app, Desktop and mobile apps, Plugins (TypeScript) |
| Regions | Not stated. | Not stated for orbs. Amp's main infrastructure is on Google Cloud in the US. |
| Facts checked | 2026-11-14 | 2026-11-14 |
OpenAI Codex cloud alternatives · Amp alternatives
Sources
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
$20 /month
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
Each cloud task runs in a VM with these default resources:
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
Plans at $100, $200, or $500 USD per month
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
Codex on the web, in the CLI, in the IDE extension, and on iOS
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
Pro plans currently have no five-hour limit. Cloud tasks may use more of your allowance than local messages.
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
No cloud-based features (GitHub code review, Slack, etc.)
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
GPT-6 Luna at Standard speed in the desktop app, subject to rollout
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
| Plus, Edu Plus | 2 | 8 GiB | 8 GiB |
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
| Pro, Business, Enterprise | 4 | 16 GiB | 32 GiB |
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
By default, a task's saved VM state is recoverable for up to seven days after you last start a turn or resume the task.
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
Larger VMs and custom specifications are available for Enterprise.
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
Publishing captures the prepared filesystem for new tasks.
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/pricing, checked 2026-11-14:
Ultrafast mode access on Pro $500
- OpenAI Codex cloud: https://learn.chatgpt.com/docs/environments/cloud-environments, checked 2026-11-14:
|Edu, Edu Pro |4 |16 GiB |32 GiB |
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
a1.small | 2 | 4 GB | $0.17/hour
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
a1.3xlarge | 16 | 44 GB | $2.13/hour
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
Every current orb size includes 60 GB of disk.
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
Orb usage is billed by the minute. A paused orb costs nothing.
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
an orb pauses 5 minutes after the agent finishes or 20 minutes after your last interaction, whichever comes later
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
These are idle timeouts, not a limit on how long an orb can run.
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
a1.3xlarge is available only to Gigawatt subscribers and Enterprise workspace members and is priced higher, due to the general scarcity of compute.
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
CPUs × $0.0504/hour
- Amp: https://ampcode.com/docs/orbs/sizes-and-costs, checked 2026-11-14:
GB memory × $0.0162/hour
- Amp: https://ampcode.com/pricing, checked 2026-11-14:
45,000 minutes of orbin' time
- Amp: https://ampcode.com/pricing, checked 2026-11-14:
$20 ∕ mo.
- Amp: https://ampcode.com/docs/markdown/pricing, checked 2026-11-14:
750 hours (45,000 minutes) of a1.small orbs. Larger orbs use the hours faster.
- Amp: https://ampcode.com/docs/markdown/pricing, checked 2026-11-14:
$20 of agent usage ($10 with the education discount) for Amp-served models and tools.
- Amp: https://ampcode.com/docs/markdown/pricing, checked 2026-11-14:
| Gigawatt | $200/month | $200 of agent usage for Amp-served models and tools. After that, buy credits at list price. | 1,000 hours of a1.xxlarge orbs. |
- Amp: https://ampcode.com/pricing, checked 2026-11-14:
Students and teachers can get Megawatt for $10 per month.
- Amp: https://ampcode.com/pricing, checked 2026-11-14:
Pay-as-you-go for Amp-hosted orbs
- Amp: https://ampcode.com/pricing, checked 2026-11-14:
An Amp paid tier gives you a big discount on orbs and includes Amp credits that you can use on any model.
- Amp: https://ampcode.com/docs/pricing, checked 2026-11-14:
Included usage resets at the end of each monthly billing period and does not roll over.
- Amp: https://ampcode.com/docs/pricing, checked 2026-11-14:
Orbs use a separate usage allowance.
- Amp: https://ampcode.com/docs/orbs, checked 2026-11-14:
Each user can start a burst of 20 metered orbs.
- Amp: https://ampcode.com/security, checked 2026-11-14:
e2b provides ephemeral compute instances for Amp orbs.
- Amp: https://ampcode.com/security, checked 2026-11-14:
Amp saves a snapshot of a prepared orb for each project and orb size.
- Amp: https://ampcode.com/security, checked 2026-11-14:
A paused orb’s filesystem and memory remain stored, encrypted at rest, until the thread is deleted.
Questions
Is OpenAI Codex cloud or Amp better?
OpenAI Codex cloud scores higher on our published method: 69/100 vs 53/100 as coding agents with cloud machines. Choose OpenAI Codex cloud for chatGPT subscribers who want cloud coding tasks without paying for machines separately. Choose Amp for running many agent tasks in parallel, each in its own paused-when-idle cloud machine. Checked 2026-11-14.
How do OpenAI Codex cloud and Amp isolate code?
OpenAI Codex cloud: Each cloud task runs in its own VM. Amp: Separate sandboxed cloud machine per thread; Amp's security page says e2b provides the compute instances. The isolation technology is not stated by Amp.