Lilac

Lilac provides a hosted inference service that leverages idle enterprise GPUs to deliver high-speed model inference at competitive prices. The platform supports various models, including MiniMax M2.7, Kimi K2.6, GLM 5.1, and Gemma 4 (31B). By routing requests to GPUs that are already powered on, Lilac ensures low latency and high throughput without the overhead associated with cold starts. This operational efficiency allows Lilac to offer its services at a lower cost compared to traditional GPU providers, as users only pay for the tokens they consume, with no minimum commitments or reserved ca…