A pay-per-task GPU router for open-source AI models
Building AI apps on rented cloud GPUs can waste money when the machine sits unused for most of the day. Lexora is a that sends each request to a live GPU node that already has the requested loaded.
An app can send an API request for a model such as Llama 3.2 or FLUX.1, then receive generated text or images without renting a . The pricing idea is to pay only for the seconds when the GPU is actually generating the result.
The service aims to avoid by keeping models ready on provider nodes. It is currently being tested for free by 10 beta users while the routing is being refined.
Key points
- Lexora routes model requests to available GPU nodes instead of requiring a .
- It supports such as Llama 3.2 and FLUX.1.
- The cost model is based on the actual seconds spent generating text or images.
- Models are kept ready on provider nodes to reduce .
- The network currently has 10 free beta users testing it.