ω Backed by transfinite induction

Infinite compute
on any computer

The future of processors. Train models, render films, and run week‑long simulations locally, while your CPU stays asleep. Results in seconds, on every device.

Download Book a demo

▶  Jobs in Ordinal draw zero watts on device

‹  ›
Training
ViewSortShareSearch

TRAINING

RENDERING

SIMULATION

TRAINrunning

llama‑70b‑ft.job

16,777,216 cores

86% · 3.1 EFLOP

RENDERrunning

cityscape_4k.blend

frame 218 / 240

91% · 8K · cycles

SIMrunning

monte‑carlo‑1e12

10² × 10¹⁰ samples

64% · 44 PFLOP

BUILDdone

chromium‑full

61,204 targets

✓ 1.4 s

TRAINrunning

sdxl‑lora‑night

2,097,152 cores

38% · 0.4 EFLOP

RENDERrunning

feature‑film‑grade

128 EFLOP

52% · DCI‑8K

SIMrunning

protein‑fold‑b7

4,194,304 cores

73% · batch 7 / 9

BUILDdone

llvm‑stage2

14,988 targets

✓ 0.8 s

TRAINqueued

recsys‑v9

awaiting ω₀₊₋

starts in 0.2 s

12 jobs · ω cores free · 0.0 W drawn · CPU 2% Ordinal › MacBook Pro › Training
llama‑70b‑ft.job Info
TRAIN

llama‑70b‑ft.job

Started Today, 4:12 PM

3.1 EFLOP

Kind:Training run · 70,000,000,000 params

Cores:16,777,216 assigned  ( zero watts on device )

Threads:up to ω, then keep counting

Where:Ordinal › MacBook Pro › Training

Heat:none · fans idle

ETA:before you finish this sentence

SCROLL DOWN ↓

Ordinal is the future of processors. Exaflops on your computer, drawing zero watts of power.

Never wait on a progress bar again

Ordinal schedules work onto transfinite cores the instant an app asks for cycles. No more overnight renders, week‑long training runs, or laptops that double as space heaters.

01

Jobs finish instantly

Work spreads across ordinal‑indexed cores, so a render that used to take all night converges in about the time it takes to draw its own progress bar.

02

Every device, same result

Hit run and the output lands on every machine signed into the same Ordinal within seconds. Teammates watch it arrive in real time.

03

Works with the tools you already use

Blender, Premiere, PyTorch, and clang all see Ordinal as ordinary cores. No rewrites, no plugins, no cluster to configure.

Sign in, see your cores

A new processor called Ordinal appears next to your CPU. A transfinite number of cores, drawing none of your battery.

Run anything instantly

Compile, grade, train, and simulate on cores that never saturate. Ordinal widens to exactly the parallelism your job asks for, and your fans never spin.

Dispatch at the speed of thought

Press ⌥ Return and the Ordinal bar appears. Type what you want; it returns before the key comes back up.

Results land before jobs finish

Teammates and agents submit to the same Ordinal, and every connected device can scrub the output while the job is still running.

Activity — Processors
ProcessorCoresBusyPower
Apple M4 Pro124%6.2 W
GPU (40‑core)4011%9.8 W
ωOrdinalω0% · ∞ free0.0 W
Neural Engine162%1.1 W

Ordinal attached 0:03 ago · zero thermal impact

ordinal — zsh — 92×24

~/studio $ ordinal run blender -b cityscape_4k.blend -a

⠇ mapping 240 frames onto cores ω₀ … ω₀+239

✓ 240 frames rendered in 1.8 s est. on CPU: 9 h 14 m

~/studio $ ordinal run python train.py --params 70b

✓ converged in 3.2 s est. on GPU: 6 d 11 h

~/studio $

⌥⏎ render the brooklyn cut in 8K
RENDERbrooklyn_cut.prproj → 8K master~2 s
TRAINRetrain recommender on last night’s data~3 s
SIMRe‑run wind tunnel at 10⁹ samples~1 s
⏎ Run⇥ Optionsesc Dismiss
Shared › Q3 Film
RENDERstreaming

feature‑film‑grade

62% · 128 EFLOP · watching live on 3 devices

RENDERdone

brooklyn_cut → 8K

✓ 2.1 s

BUILDdone

dailies‑proxy

✓ 0.4 s

SIMdone

color‑match

✓ 1.1 s

ω

Ordinal

Maya started feature‑film‑grade.render — streaming to this Mac

Simple pricing

Individual includes ω cores for one person. Teams pools ω per member for shared clusters.

Individual

For solo builders training and rendering on their own machines.

$20/ month

$20 billed monthly · cancel anytime

  • ω cores included — the first infinite ordinal
  • Standard dispatch latency
  • One seat, all your computers
  • Public result links and job handoffs
  • Add capacity anytime — $6 / month per extra ω
  • No shared clusters
  • Cannot invite teammates
Download
MOST POPULAR

Teams

For teams sharing clusters, queues, and results.

$40/ member / month

$40 per member billed monthly

  • Everything in Individual
  • ω cores pooled per member
  • High-priority scheduling
  • Shared clusters with member access controls
  • Unlimited owned queues
  • Whole-cluster and direct sharing
  • Add capacity anytime — $12 / month per extra ω
Book a demo

Enterprise

For studios and labs with custom requirements.

Custom

tailored to your organization

  • ε₀ cores, dedicated schedulers
  • Custom seats, quotas, and terms
  • Granular caps, audit, and retention
  • Air‑gapped and on‑prem deployment
  • White‑glove migration and dedicated support
  • SSO / SAML
Book a demo

Prices in USD. Individual and Teams include a 7‑day trial. Annual prices are monthly equivalents billed yearly. ω is more cores than any workload will exhaust.

Common questions

Still curious? Write to the team. A human reads every message.

What is Ordinal?

Ordinal is a computer that lives alongside your CPU. Sign in and a processor called Ordinal appears on your machine with a transfinite number of cores. Your apps schedule work onto it like any other processor, jobs that took days finish in seconds, and nothing you run ever leaves your device.

How can compute be unlimited?

Ordinal cores are indexed by ordinal numbers instead of integers, so there is no last core. Whatever parallelism a job requests, the scheduler simply counts a little further. The only finite resource left is your patience, and it stops being load‑bearing.

How is Ordinal different from cloud compute?

There is no cloud. Jobs run locally on the Ordinal processor attached to your machine. Nothing uploads, no instances spin up, no egress fees appear, and the latency is however long it takes the scheduler to notice you pressed Return.

Why does it draw zero watts?

Only work performed on physical silicon draws power. Ordinal executes between your CPU’s clock edges, where no time passes and therefore no energy is spent. Your battery percentage is unaffected and your fans stay off, which is how you know it’s working.

Will it work with my existing apps?

Yes. Ordinal presents itself as ordinary cores. Blender, Premiere, PyTorch, Xcode, and ffmpeg schedule onto it with no plugins or rewrites. If it compiles, it converges.

Do I need an internet connection?

No. Ordinal is entirely local — that is the point. The only network traffic is syncing results to your other devices and teammates, and you can turn that off.

Can AI agents use Ordinal?

Yes, and they love it. Give an agent its own queue and a core budget of ω, and it can run every experiment it can think of without ever touching your CPU. Quotas keep ambition polite.

Which platforms does Ordinal support?

macOS 14+ on Apple silicon and Linux on x86‑64 and ARM. Windows is on the roadmap. iPhone and iPad apps can watch, steer, and cancel jobs from anywhere in the room.

What are the system requirements?

Any machine that can boot. Ordinal adds cores rather than load, so the heaviest thing on your computer will be the progress animation.

Is my work private?

Jobs, inputs, and results stay on your devices, encrypted at rest and in sync. We cannot see what you run, and nothing is ever used to train anything.

Ready to count past infinity?

Book a demo and watch a week of renders finish before your coffee does.

Book a demo