Models
Simo-1: real-time judgment
Simo-1 (simo-1) is the default. It makes real-time judgments for agents, pipelines and per-event processing, answering one question over a screenshot in 64 ms.
Why it exists
Most judgment calls in software are routine, and they happen constantly: every agent step, every ticket, every event in a stream. Simo-1 is built for that volume. It keeps the per-call cost low enough to put a judgment on every step, rather than only on the few you can afford.
Published latency
Measured on one request over one 1280×720 screenshot, questions answered in the same pass.
| Questions | Simo-1 latency |
|---|---|
| 1 | 64 ms |
| 10 | 104 ms |
| Each added question | ~10 ms |
Because the situation is read once, asking ten questions instead of one adds about 40 ms.
What it is good at
- Agent supervision: done-ness, next action and risk in one call per step.
- Ticket and inbox triage: route, urgency, sentiment and order ID in one request.
- Visual QA assertions and moderation at stream volume.
- LLM judging on every response.
When to move up to Simo-1 Pro
Choose Simo-1 Pro for high-stakes gating, hard calls, and pointing at the exact spot on screen.
Frequently asked questions
How fast is Simo-1?
64 ms for one question and 104 ms for ten questions over one 1280×720 screenshot, about 10 ms per added question.
What is Simo-1 for?
Real-time judgments for agents, pipelines and per-event processing. It is the default Simo model.