Answers before you begin.
Practical answers about accounts, measurements, imports, exports and the current workspace.
What can I use today?
Create an account in the benchmark workspace, save measured latency samples, import run JSON, compare two runs, and export CSV or JSON.
How do I collect measurements?
Download the benchmark CLI and run it locally against an OpenAI-compatible chat-completions endpoint you control or are authorized to use. See the quickstart.
Does Orbitriva need my provider key?
No provider key is needed to save benchmark records. The local CLI can read MODEL_API_KEY from your own environment; never upload that key or include it in a run file.
Are dashboard examples my account data?
The public visualizations use sample data. Your actual measurements are stored separately in your private workspace and are not automatically replaced by the public examples.
How are latency metrics calculated?
Percentiles use nearest rank over submitted samples. Output throughput uses total reported output tokens divided by total measured seconds and is unavailable if token usage is missing.
Can I delete or export my data?
Yes. Export CSV from the workspace or JSON from the account dialog. Delete individual runs, or delete your account and its active database records. Deletion cannot be undone.
Is email verification or password recovery available?
Not in this developer release. Use a unique password and keep it safe. For account assistance, contact support; never send a password.
How do I discuss commercial use?
Use pricing and deployment enquiries. No payment details are collected by this site and no enterprise entitlement is implied by creating an account.