Pricing
Prepaid tokens, no seats, no minimums. New accounts start with 500k free tokens — enough for a few QC jobs or a dozen lookups.
Token packs
$25
One-time payment. Tokens never expire while the account is active.
Request beta access$100
One-time payment. Tokens never expire while the account is active.
Request beta access$500
One-time payment. Tokens never expire while the account is active.
Request beta accessInvoice billing, dedicated compute workspace, self-hosted sandbox option, DPA, SSO, priority support.
Talk to usWhat a job costs
A job uses tokens in proportion to what the agent reads, computes and writes: model calls, tool use and sandbox time all count. There are no per-job caps: the right amount of work for a scientific task is not knowable up front, so a running job is never stopped for budget. Its usage is debited when it ends; a balance can dip slightly negative and blocks the next submit until topped up.
| job | what runs | class | wall time | tokens |
|---|---|---|---|---|
| CRO assay QC | Recompute plate statistics, refit IC50 curves, flag anomalies in a CRO xlsx deliverable | specialist (standard model) | ≈ 11 min | 185k |
| Database lookup | One structured question against a public database (GTEx, UniProt, Open Targets, ChEMBL, DepMap, PubMed …), with a figure | lookup (standard model) | ≈ 2 min | 58k |
| Research phylogeny | Family selection, MAFFT + FastTree, ancestral reconstruction, structural superposition, 5 figures | research (frontier model) | ≈ 39 min | 1.11M |
Measured on the production engine with the datasets shown under Examples. The billing block on every result shows the tokens debited and your balance after.
Job classes
| class | model tier | wall-clock backstop | intended for |
|---|---|---|---|
specialist | standard model | 60 min | assay QC, curve fits, plate statistics, PK parameters, plots over supplied files |
lookup | standard model | 30 min | database / literature questions: DepMap, Open Targets, ChEMBL, UniProt, PubMed, trials |
research | frontier model | 180 min | open questions that need planning, many tools and synthesis (phylogenies, dossiers) |
Classes set the model tier and a wall-clock backstop only.
Limits
- 2 jobs in flight per account by default (raise it on request); 60 requests/min, 10 submits/min per key.
- Inputs: inline ≤ 20 MB, https URL fetch ≤ 500 MB, multipart upload ≤ 600 MB.
- Sandbox: Ubuntu 24.04, ≤ 8 GB RAM, ≤ 10 GB disk, no GPU.
- Artifacts are retained 90 days; the result document (summary, findings, billing) is kept indefinitely.
FAQ
Do tokens expire?
No, as long as the account is active. Used tokens are non-refundable; see the terms.
What happens to the data I send?
It stays yours and stays confidential. Prompts, input files and everything a job produces are never used to train any model, never sold or shared, and never mixed between customers: each job runs in its own sandbox that is destroyed when the job ends, inputs are mounted read-only and the upload is deleted once mounted. Our staff do not read your content in the ordinary course of running the service. Artifacts are kept 90 days and you can delete a job or your whole account at any time. Details in the privacy policy; we sign a DPA on request.
Can I get an invoice instead of paying by card?
Yes, on the enterprise tier. Email ops@scientistworkbench.com.
Is there a free tier?
Every new account gets 500k tokens. There is no ongoing free allowance.
Can ScientistWorkbench run HPC jobs like AlphaFold?
No. Jobs run on a single sandbox with no GPU, which rules out structure prediction and other HPC workloads. We hope to add these capabilities in the future.
What if a job fails?
Jobs that fail before running anything use no tokens. Jobs that are cancelled or hit the backstop are debited for the work done and still return whatever was saved. If our side broke a job, tell us and we credit it back.