Idle CPOffersHow it worksFAQSuppliersFRWaitlist

Batch AI compute

Batch AI compute, at off-peak hours.

Submit a batch of transcription, embeddings or evaluations. Choose Guaranteed to start right away, or Flexible to pay less and get the result within 6 to 24 hours.

Service in testing. Early sign-ups get a preferential rate at launch.

Example command, the service is not open yet
$ idlecp batch send october-audio.zip --tier flexible
3,412 files received
✓ Batch accepted, delivery estimated within 18 h
! 12 unreadable files, skipped

Two tiers

One batch, two ways to run it

You pick the tier batch by batch. Flexible costs less because we place it on capacity that is free outside peak hours.

Guaranteed

Immediate start

The batch starts right away on reserved capacity. The deadline is kept.

  • Sensitive data accepted
  • Deadline stated in the quote, then kept
  • Reference price

Flexible

Delivered within 6 to 24 h

The batch waits for a window of free capacity. You get the result within the stated window, with no guaranteed hour.

  • Non-sensitive data only
  • Delivery window of 6 to 24 h
  • Lower price than Guaranteed
CriterionGuaranteedFlexible
StartImmediateWithin the 6 to 24 h window
Delivery timeStated and keptNot guaranteed
Sensitive dataAcceptedRefused
PriceReferenceLower

Prices are given by quote, per batch. No rate is published before launch.

How it works

From upload to result in three steps

  1. Submit the batch

    You send your files and describe the job: transcription, embeddings, model evaluation.

  2. Choose the tier

    You get a quote for Guaranteed and for Flexible, with the deadline of each. You pick one.

  3. Get the result

    We spread the batch across available capacity, then deliver the results with an execution report.

Use cases

Jobs that do not need the result this second

  • Bulk audio transcriptionThousands of meeting or call recordings, transcribed overnight.
  • Corpus embeddingsIndex documents for internal search or an assistant without tying up your servers.
  • Model evaluationReplay a full test set on several model versions before going to production.
  • Document processingExtraction, sorting or summaries of scanned files, in batches of several thousand pages.

Waitlist

Get the preferential launch rate

Tell us what you process and how much. We write to you before opening, and we use your answers to decide what to build first.

An estimate is enough.

Tier you are interested in

FAQ

What people ask us most

When does the service open?

We have no firm date. People on the list are told first and can test before general opening.

Can my sensitive data run on Flexible?

No. Sensitive data runs on Guaranteed only. Flexible is for non-sensitive data.

What capacity do my batches run on?

On unused compute capacity: servers from partner providers and, later, machines from volunteer individuals and companies. Guaranteed does not rely on individuals' machines.

How much does it cost?

The price depends on the batch and the tier. You get a quote before starting the job. Early sign-ups get a preferential rate, with terms stated before launch.