Daytona
On-demand and preemptible GPU sandboxes for inference, fine-tuning and AI agents, billed per second.
New York, United States Founded 2023 5 regions 9 GPU modelsAt a glance
prices checked 5h ago- Cheapest GPU
- $0.99/hr
- NVIDIA GeForce RTX 4090 24GB
- Fastest GPU
- $5.99/hr
- AMD Instinct MI355X OAM
- GPU models
- 9
- 18 configurations
- Regions
- 5
What's good about Daytona
- Exclusive GPU allocation One sandbox per GPU, never shared
- SDKs in five languages Python, TypeScript, Ruby, Go, Java
- Built for untrusted code Isolated sandboxes, SOC 2 Type II
- Freeform resources up to 16 vCPU, 192 GB RAM, 512 GB disk per GPU
- Capacity API and fallbacks Live GPU availability, ordered type lists
- Billing Per second, no minimum
Available GPUs
Cheapest first, per GPU per hour.
| GPU | Configuration | Memory | Price | Region | |
|---|---|---|---|---|---|
| AMD Instinct MI355X OAM CDNA 4 | 1 GPU Spot | 288 GB | $3.44/hr | - | Rent |
| AMD Instinct MI355X OAM CDNA 4 | 1 GPU | 288 GB | $5.99/hr | - | Rent |
| NVIDIA B200 SXM 180GB Blackwell | 1 GPU Spot | 180 GB | $3.59/hr | - | Rent |
| NVIDIA B300 SXM 262GB Blackwell Ultra | 1 GPU Spot | 262.5 GB | $4.08/hr | - | Rent |
| NVIDIA B200 SXM 180GB Blackwell | 1 GPU | 180 GB | $6.25/hr | - | Rent |
| NVIDIA B300 SXM 262GB Blackwell Ultra | 1 GPU | 262.5 GB | $7.10/hr | - | Rent |
| NVIDIA H100 SXM5 80GB Hopper | 1 GPU Spot | 80 GB | $2.27/hr | - | Rent |
| NVIDIA H200 SXM 141GB Hopper | 1 GPU Spot | 141 GB | $2.61/hr | - | Rent |
| NVIDIA H100 SXM5 80GB Hopper | 1 GPU | 80 GB | $3.95/hr | - | Rent |
| NVIDIA H200 SXM 141GB Hopper | 1 GPU | 141 GB | $4.54/hr | - | Rent |
Daytona head to head
Compare any two
About Daytona
Daytona is a cloud GPU platform built around isolated sandboxes rather than virtual machines. A sandbox starts from any Docker or OCI image and is created from the dashboard, the REST API, the CLI or an SDK in Python, TypeScript, Ruby, Go or Java. Each one can carry one to eight GPUs with exclusive device allocation: an assigned GPU belongs to a single sandbox for as long as it runs and is never shared.
The fleet covers the NVIDIA B300, H100, H200, RTX PRO 6000, RTX 5090 and RTX 4090 and the AMD Instinct MI355X, with the NVIDIA B200 on the rate card ahead of launch. Every GPU type is sold on-demand and as preemptible capacity, billed per second with no minimum commitment, and reserved capacity is available under enterprise agreements. vCPU, memory and disk are configured per sandbox and metered separately, up to 16 vCPUs, 192 GB of RAM and 512 GB of disk per GPU.
Around the sandbox itself, snapshots make an environment reusable in seconds, volumes hold datasets and checkpoints across sandboxes, and a live capacity API with ordered GPU-type fallback lists lets a job take the first available card rather than wait for a specific one.
Founded in 2023 by the team behind Codeanywhere and headquartered in New York, Daytona is SOC 2 Type II audited. The platform is positioned for LLM inference and serving, fine-tuning, model evaluations and the code execution behind AI agents, where running untrusted or model-generated code safely is the point. GPU capacity is available across the United States, Canada, the United Kingdom, the European Union and Norway.
- One sandbox per GPU, never shared Exclusive GPU allocation
- An assigned GPU device belongs to a single sandbox for as long as it runs. Up to eight GPUs per sandbox, with no time-slicing or shared-tenancy behind the rate.
- Python, TypeScript, Ruby, Go, Java SDKs in five languages
- A GPU sandbox is a few lines of code in any of the SDKs, or a call to the REST API, the CLI or the dashboard. Every sandbox starts from any Docker or OCI image, so an existing container runs unchanged.
- Isolated sandboxes, SOC 2 Type II Built for untrusted code
- The platform is designed for running AI-generated and untrusted code safely. Snapshots make an environment reusable in seconds and volumes carry datasets and checkpoints between sandboxes.
- up to 16 vCPU, 192 GB RAM, 512 GB disk per GPU Freeform resources
- vCPU, memory and disk are set per sandbox rather than fixed to an instance shape, and metered separately from the GPU. The caps scale with the number of GPUs attached.
- Live GPU availability, ordered type lists Capacity API and fallbacks
- A capacity endpoint reports free on-demand and preemptible units per GPU type, and a sandbox request can name an ordered list of acceptable types so a job takes the first available card.
- Per second, no minimum Billing
- On-demand and preemptible rates for every GPU type, billed per second from the moment a sandbox starts with no minimum commitment. Reserved capacity is quoted under enterprise agreements.
Pricing options
Services
Compliance and certifications
Independently verified standards governing security, privacy and operations.
Regional coverage
Where workloads run: primary datacenters and active deployment zones.
- United States Primary
- Canada
- United Kingdom
- European Union
- Norway
Frequently asked questions
Common questions about working with Daytona.
Which GPUs does Daytona offer?
NVIDIA B300, H100, H200, RTX PRO 6000, RTX 5090 and RTX 4090, and the AMD Instinct MI355X, each available as a sandbox with one to eight GPUs.
What is the difference between on-demand and preemptible?
On-demand sandboxes keep their GPUs until you stop them. Preemptible sandboxes use spare capacity at a lower per-hour rate and can be reclaimed by Daytona when on-demand demand needs the GPUs, so they suit fault-tolerant work such as batch inference, evaluations and checkpointed training. Both are billed per second. Reserved capacity is available under enterprise agreements on request.
How does Daytona billing work?
Per second, with no minimum commitment. The GPU rate covers the GPU only; vCPU, memory and disk are configured per sandbox and metered separately per hour. There is no subscription and no minimum spend, and new accounts start with free compute credit.
Where do Daytona GPU sandboxes run?
Daytona has GPU capacity in the United States, Canada, the United Kingdom, the European Union and Norway. GPU sandboxes are scheduled onto an aggregated shared region that spans that capacity, so a request takes the first available GPU of the type you ask for rather than targeting a single site. Dedicated and custom regions are available for organisations that need a fixed location.
How do I create a GPU sandbox?
From the dashboard, the REST API, the CLI, or the SDKs for Python, TypeScript, Ruby, Go and Java. You pick a Docker or OCI image, the number of GPUs and a GPU type, or an ordered list of acceptable types so the sandbox falls back to an alternative when the first choice is unavailable. A capacity endpoint reports live on-demand and preemptible availability per GPU type before you create one.
Is Daytona suitable for running AI-generated code?
That is what the platform is built for. Each sandbox is isolated, starts from an image you control, and can be snapshotted and reused, so agent code execution, evaluations and model serving run without access to your own systems. Daytona is SOC 2 Type II audited, with the report available at trust.daytona.io.
Ready to deploy on Daytona?
Visit Daytona