Skip to content

Your overnight AI jobs shouldn’t cost daytime prices.

Otium AI runs multi-step AI pipelines on spare cloud GPUs when they are cheapest. Send every step at once, with one deadline for the whole job, and the same work costs far less.

For teams and researchers with a backlog of AI work: enriching a catalog, extracting from documents, labelling a research corpus, summarising overnight.

Free to try on our demo model. No credit card, no spam: we email you a link to finish.

Or try a pipeline in the playground, no account needed →

Example: enriching 100,000 product listings costs $28 at OpenAI’s regular price, $14 through its batch service, and $11.42 on Otium with a week’s deadline: 59% less. See the prices

Workflows · graph engineering

One deadline for the whole job, not one per step

Plenty of AI work runs in steps: extract, then classify, then summarise. Send each step as its own batch and you wait for each one in turn, so a week per step becomes seven weeks for seven steps. Send all the steps together as one workflow and Otium sees the whole graph, a DAG of your steps: one deadline covers the whole job.

Seven steps, sent as seven batches
up to 7 weeks
a week’s window each, one after another
The same seven steps, as one workflow
1 week
one window for the whole job
  • The order comes from your prompts: a step that reads another’s output runs after it.
  • Every output is checked against a schema before the next step sees it.
  • A step can call your own API, or wait for a person, mid-run.
Compile one in your browser → Read the spec
modelextractotium-smallclassifyotium-smalldescribeotium-medium
Describe this graph in words
  1. extract — otium-small, on the critical path. Starts the graph.
  2. classify — otium-small. Runs after extract.
  3. describe — otium-medium, on the critical path. Runs after extract, across a model change barrier.

Three steps, and nothing says what runs when: classify and describe both read extract, so they run side by side once it finishes.

The document it compiles from
workflow.yaml
name: enrich-listing
description: Extract the facts, then classify and describe in parallel.

input:
  schema:
    type: object
    required: [listing]
    properties:
      listing: { type: string }

schemas:
  facts:
    type: object
    required: [brand, specs]
    properties:
      brand: { type: string }
      specs: { type: string }

nodes:
  extract:
    target: { product: otium-small }
    estimate: { tokens_in: 900, tokens_out: 200 }
    body:
      messages:
        - role: user
          content: "List the brand and specs as JSON: ${input.listing}"
    output:
      validators:
        - { type: json_schema, schema: { $ref: "#/schemas/facts" } }
      on_invalid: { retry: 2 }

  classify:
    target: { product: otium-small }
    estimate: { tokens_in: 300, tokens_out: 20 }
    body:
      messages:
        - role: user
          content: "Pick one category for: ${nodes.extract.output.specs}"

  describe:
    target: { product: otium-medium }
    estimate: { tokens_in: 300, tokens_out: 250 }
    body:
      messages:
        - role: user
          content: "Describe it: ${nodes.extract.output.specs}"

How it works

  1. 1

    Change one line of code

    Point the OpenAI client you already use at Otium.

  2. 2

    Send it all, with a deadline

    The whole backlog, or a whole pipeline, and when you need it: within a day, or a week.

  3. 3

    Collect your results

    They arrive by your deadline. 30 days later, your data is deleted.

The longer you can wait, the less you pay

Two prices per size, and the deadline is the only dial. Repeated input is free.

See all prices

The GPU prices under ours are public: live GPU prices and the monthly market report.

Enrich 100,000 product listings
800 input + 400 output tokens each: 80M in, 40M out. Small class.
Interactive API, list price
gpt-6-luna standard, answers in seconds
$28.00
Provider batch API
same model, fixed 24-hour window
$14.00
Otium Standard
< 24 hours, open-weight model
$13.44
Otium Remnant
< 7 days: the same work, more patience
$11.42

Built for work you would rather not hand to an AI company

Who it’s for

Stop paying for speed you don’t need.

Otium is in beta. Leave your email and we’ll send you a link to create your account. Then point your OpenAI-compatible client at our base URL and send your first batch.

Free to try on our demo model. No credit card, no spam: we email you a link to finish.

Otium is in beta and open to everyone. Already have an account? Sign in.