High‑end computers in New York, for agents across the Americas
United States · High‑end runs your Uno computer on the latest generation of AMD Ryzen processors, DDR5 memory and NVMe disks. On Dax, the open sandbox benchmark, it clones, installs and typechecks a real project in about 40 seconds.
Right now New York is fully booked. New computers for the Americas run in the Netherlands, and your agent goes online from a clean IP in the United States.
- Dax, whole job
- about 40 s
- Processor
- Latest AMD Ryzen
- Serves
- US, Canada, Latin America
The fastest servers we run
High‑end is a different machine, not a setting: the same Uno computer, with the same files, tools and agent, on hardware picked for speed.
The latest AMD Ryzen
Typechecks, compiles and test runs wait on the processor. This region runs on the newest generation of AMD Ryzen, the quickest we have put under an Uno computer, with DDR5 memory.
NVMe disks
An install writes thousands of small files. They land on mirrored NVMe disks, so a fresh install does not wait on storage.
Ready for builds
Compilers and build tools are already in the image, and packages come from a mirror in the US. Your agent starts on the work, not on setup.
Dax: about 40 seconds from an empty computer to a checked project
Dax is the job an agent does before it can start coding: set up the toolchain, clone OpenCode, install about 950 MB of packages, typecheck the whole repository. A new computer every run, so nothing is warm.
Where the 40.6 seconds go
Seconds, medians of our runs. Typecheck is the processor; install is the network and the disk.
Show as a table
| Option | s | Shape |
|---|---|---|
| Typecheck | 21.1 | The compiler reads and checks every file of the monorepo |
| Install packages | 11 | About 950 MB of packages from a US mirror |
| Clean-up and start | 3.4 | The job deletes its 3 GB working folder and runs its own checks. The call itself, sent straight to the server in New York, takes about 0.1 s |
| Prepare | 2.9 | Build tools are already in the image, packages come from a US mirror |
| Clone | 1.3 | git clone of the OpenCode repository |
| Download Bun | 0.9 | The Bun runtime the job installs with |
Dax, whole job, next to the public leaderboard
Seconds, lower is better. Grey: the ComputeSDK leaderboard of 9 October, top 5. Violet: our own run of the same script; we are not on the leaderboard yet.
Show as a table
| Option | s | Shape |
|---|---|---|
| Isorun | 34.5 | Leaderboard, 9 October 2026 |
| Namespace | 40.1 | Leaderboard, 9 October 2026 |
| Uno, US · High‑end | 40.6 | Our run through our public API, 19 October 2026. Median of 5 runs. Not on the leaderboard yet |
| Gravix | 41.7 | Leaderboard, 9 October 2026 |
| GiveMeANode | 41.7 | Leaderboard, 9 October 2026 |
| Blaxel | 45.9 | Leaderboard, 9 October 2026 |
Leaderboard of 9 October 2026: 21 providers, one run per provider. Our number is the median of 5 runs on 8 vCPU · 16 GB, a fresh computer every run, timed through our public API, the way the ComputeSDK runner calls us.
The official result goes here
ComputeSDK runs Dax on every listed provider each week and publishes the times. When our row appears, this page shows the official time and links to it, whatever it says.
Prices in the US region
Pro at $90 includes both regions, so you can run computers in New York and in the Netherlands on one plan.
| Plan | United States · High‑end | With Uno AI | Netherlands |
|---|---|---|---|
| Small1 vCPU · 1 GB | $10 | $25 | $5 |
| Plus4 vCPU · 4 GB | $30 | $60 | $20 |
| Pro16 vCPU · 16 GB | $90both regions included | $140 | $70+ $20 a month adds High‑end |
| Max32 vCPU · 48 GB | $200both regions | $300 | $200both regions |
- Prices per month, + VAT. High‑end is billed monthly for now.
- Signed up from the Americas? These are your prices, and Pro covers both regions.
- Anywhere else, pick High‑end when you buy Small or Plus, or tick it on Pro.
What High‑end does not change
- Memory in this region has no error correction (ECC).
- vCPU are shared. With busy neighbours a run takes longer than ours did.
- Dax ran on 8 vCPU · 16 GB. Plus boosts to 8 vCPU and Pro has 16, so your size decides your time.
- A benchmark is one job. Work that leans on the processor gains the most; a job that waits on someone else's server does not.