Smart AI
You ask; we choose. Routing weighs the task, your plan and cost, and the answer says which model wrote it, how long it took to the first token, and what it spent. If a provider fails, it fails over and tells you it did.
Compare
Send one prompt to up to six models at once. They stream together with latency and cost per column. Pick the best answer, or have them synthesised into one. The estimate is on screen before anything runs.
AI Council
Several answering models reply independently, in different roles. One model lines up their claims; another rules and writes a verdict with reasoning, risks and what would change its mind. The confidence figure is a weighted heuristic, and you can open it and see every term.
Deep Research
A worker plans searches, reads the pages, cross-checks what they claim against each other and writes a report with numbered citations. Contradictions between sources are listed rather than smoothed over, and sentences nothing supports are flagged. The whole run is priced before it starts.
A plan gives you AI Credits each month. Every run is priced before it starts, the exact cost is shown after it finishes, and whatever a run does not use is returned to your balance.
Consensus is not truth
Several models agreeing makes an answer better supported, not correct. The Council's confidence score is a product heuristic with published components, and the interface says so wherever it appears.
A plan here is not their subscription
Access to a model through superai is not the same product as that provider's own consumer subscription, and we do not imply it is.
No surprise spending
Compare, Council and Deep Research price the whole run before it starts and refuse to start what you cannot afford. Nothing here charges you for an answer you never saw.