Three ways to work with us.
Two start with an email and a call about the work you want measured or trained; we scope those from there, and there is no price list. The third you can start on your own, today.
Paid evaluation
You bring the tasks: a few hundred from your own production work. We run every candidate model and configuration on them under a hard budget and report what each one completes and what it costs per completed task, graded by code.
You get · A report, with every case we could not grade listed. Your cases are never published, never sold, and never go into anything we sell.
Verified EvalCustom environments
Built to order for a task family you want to train: the agent has a budget and has to pay for what the task needs. The reward comes from the settlement ledger, not a judge.
You get · The environment, the task set, the graders, reviewed golden trajectories, a calibration report and a statement of rights.
EnvironmentsDeveloper API
Pay per call, with no call to book: one endpoint, each call priced before it runs. Pay from a wallet with no account, or get an API key and pay by card.
For developersWhat we never do
- Sell or license customer data: prompts, outputs, wallets or transactions.
- Build our own models. We do have one agent of our own, Franklin, and it is open source.
How it starts
Write to hello@blockrun.ai with what you want measured or trained, and for whom. We reply within two business days.