Token-Max · coding plans
Flat monthly plans sized for real agent burn — about $0.02 per 1M tokens on every tier — on GPU cells we own and tune ourselves. Warm-cache fast, OpenAI-compatible, metered live. Capacity is real and finite — so plans are application-gated.
2.1s
788K-token warm resend
3.09s
1M-token warm resend
280–330ms
Warm interactive turns
$0.15/1M in
Flash list price
Solo
$49/mo
100M tokens / day
≈ 3B tokens a month
One heavy agent seat — your daily driver
Team
$199/mo
300M tokens / day
≈ 9B tokens a month
A small team or parallel agent fleets
Max1B/day
$499/mo
1B tokens / day
≈ 30B tokens a month
Agent fleets and long-horizon runs — a billion-token day
Founding cohort: your rate locks for 12 months. Budgets reset daily (no rollover) and are ceilings, not forecasts — Max seats are capacity-booked: we reserve dedicated cell time for your workload. Need more mid-day? Top up pay-as-you-go credits at list price; plans never block a run.
We run dedicated cells — H100 and RTX PRO 6000 boxes we kernel-tune ourselves — not a broker's dashboard. Tokens per day is a physical number here, so seats go out in small groups as capacity comes online. Applications keep plan quality high for the builders already on the cells — and your place in the queue is shown the moment you apply.
Apply in two minutes — approved plans start the same week.
OpenAI-compatible endpoint today: point Cursor, Cline, Aider, Continue, or Zed at runaii.cloud and keep your workflow. Live token meter in the console, cached reads priced at a fraction of list.
Claude Code speaks Anthropic-format — that endpoint is in active build. Flag it in your application and you're first in line when it lands.
Prefer pay-as-you-go? $5 free credits, no application → · Early-stage startup? Get $500 in credits →