Priority compute
Spend credits to guarantee your workflow runs even when the network is busy.
Completely free access
Credits come only from contribution.
- Contribute verified compute while idle to earn credits
- Credits give priority scheduling; best‑effort exists when idle
- No subscriptions, no purchases, no crypto; credits aren’t transferable
- Contribute across many model sizes (1.2B–120B+)
The simple rules
- Contribute verified work while idle → credits are saved (the only way to earn them).
- Spend credits → priority scheduling for parallel workloads.
- No contribution? You can still use best‑effort access when idle capacity exists (limited + preemptible).
Fairness: contribution is weighted by recent reliability, so “contribute once, use forever”
doesn’t work. Credits are not sold or transferred.
How credits are used for compute
Credits are a direct budget for verified work, priced by network status.
Parallel workload budget
Allocate credits across subtasks (drafts, checks, translations) and the scheduler fans out for you.
Verified‑token accounting
Credits are tied to verified work. Earned credits only come from contribution.
Pricing model (simple): credits spent ≈ verified tokens ÷ 1,000 × network multiplier.
Multipliers reflect demand (example: idle 0.6×, normal 1.0×, busy 1.4×).
Credits savings estimator (rough)
Estimate what you can earn, then spend for priority compute when demand is high.
Selected: 4.0 hours/day
Estimated verified work
—
Credits saved
—
Note
—
Community work pool: you can also reward governance/dev/mod work via an expiring voucher
pool (example target: 1% of delivered compute per epoch) without inflating the main credits.