Auto-research for vision language models
Send a video stream in real time and a question about it. Easy tasks go to a small, fast model. Hard ones go to a frontier model. You send one request and never pick between them.
Request early access below.
Easy, medium, hard. One endpoint.
OpenVector chooses the best & cheapest vision model for any task.
Use smaller models for straightforward tasks
Reading a clear label or checking whether a door is shut may not require a frontier model. Our router is designed to send straightforward tasks to smaller models, reducing unnecessary processing.
Hard work does need one
A cluttered scene, an ambiguous question, a judgement call about whether something is unsafe. The router sends these up to a frontier model.
Keep the same API request
Your application uses one endpoint and request format. We manage model selection behind the API, so you do not need to build separate integrations for each model.