What you get
Send the work, say when you need it, and Otium runs it on spare cloud GPUs before the deadline. Your code changes by one line.
Workflows
A multi-step pipeline sent as one document, with one deadline for the whole job. Otium works out the graph, a DAG, from which step reads which output, so you write prompts, not plumbing: graph engineering with the graph derived for you. No other batch service takes a pipeline.
- The workflow document
Every field of a workflow, how a step reads another step’s output, and how the order is worked out from that.
- Conditions
Run a step only when the data calls for it, and include part of a prompt only when it applies.
- Validation and fallback
Check every step’s output against a schema, retry it, hand it to a bigger model, and see what happened.
- Assets
Reference data every item in a batch shares, published once, pinned by each batch, and kept out of your requests.
- Tools and external steps
Call your own HTTP endpoint from a step, pause a run while your system or a person does the work, and get notified when a batch finishes.
- Running workflows
Publish, submit a batch, choose a window, follow runs, collect outputs, and know what it costs.
- Testing
Unit tests with canned step outputs, prompt snapshots, and the compiler’s checks, run offline and in CI.
- Try it in your browser
The real compiler, in the page: write a workflow or load an example, and see its graph, timing and cost.
A pipeline as one job
One workflow for a whole pipeline →
Every step of a multi-step job as one document with one deadline. Otium derives the graph, runs independent steps side by side, checks each output, and prices the job before it runs.
Otium is your queue →
Hand us the whole backlog, as a batch file of up to 200 MB or one workflow run. We hold it, retry what fails, and deliver by the deadline.
Works with the OpenAI client you have
Change the address your client sends to and keep your code, plus one setting for how long you can wait.
Pick a size, not a model name
Small or medium today, large to come, each on a good current open model. Few models is part of what keeps the price low.
Pay less, plainly
Your deadline sets the price
The longer you can wait, the more time we have to find cheap spare GPUs in whichever region has them, and the less you pay.
Repeated input is free →
When a prompt starts with text we have just processed, that part costs nothing. No cached-input rate, no charge to write the cache.
The market under the price, published →
Live spot and on-demand GPU prices by machine and region, and a monthly report, so you can check the discount yourself.
A spending cap that holds
Set a monthly limit and new work waits at it instead of billing past it. Prepaid, with optional top-ups.
See what every job cost
The cost of every job, and your spending day by day, in your dashboard.
Your data, your model
Deleted on schedule, never trained on
Inputs and results are deleted 30 days after your job finishes, never logged, and never used to train a model.
Your data, your own key
Each customer’s data is encrypted with its own key, so a mistake in our access controls shows nobody anything.
Check our claims in the code
The code that receives your requests and the code that encrypts your data are open source.
Send us all of it, at once
Otium is your queue, not just a model. Hand us the whole backlog in one go with a deadline, and let us do the scheduling.
-
More of your input is free
Requests that share instructions are served from cache once the first has run, and cached input costs nothing. The more of them arrive together, the more of your bill disappears.
-
Big batches start sooner
We wait until there is enough work to fill a machine before starting one. A whole backlog fills it at once; a trickle of requests waits to be gathered.
-
No queue of your own to run
No rate limiter, no retry loop, no drip-feeding a backlog through the day. Upload it as one batch file (up to 200 MB) or one workflow run: we hold it safely, retry what fails, and deliver it by the deadline.
The lifecycle of a job
- 01
Change one line of code
Point the OpenAI client you already use at Otium. Same requests, same code, nothing new to learn.
- 02
Send it all, with a deadline
Send the whole backlog at once and say when you need it: within a day, or within a week. The longer you can wait, the less you pay. If your work has several steps, send them together and one deadline covers all of them.
- 03
We run it when computers are cheap
We hold your job safely and run it on spare cloud computers when they are cheapest. If a machine is taken away mid-job, we restart the unfinished part. You can check progress at any time.
- 04
Get your results; then we delete your data
Your results arrive by your deadline and stay available for 30 days so you can collect them. Then your inputs and results are deleted. We keep only what we need to bill you, never the content.
Point your client at our base URL.
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.