# Lilac

> Lilac is the GPU cloud built for fast-moving startups, with dedicated GPU infrastructure, OpenAI-compatible serverless inference, and flexible options for committed capacity.

## Key Pages

- [Homepage](https://getlilac.com/): The GPU cloud built for fast-moving startups.
- [Dedicated GPUs](https://getlilac.com/dedicated-gpus): Dedicated GPU virtual machines, bare metal, and clusters sourced and managed through Lilac.
- [Serverless Inference](https://getlilac.com/models): OpenAI-compatible per-token inference without a GPU reservation.
- [Subscriptions](https://getlilac.com/#subscriptions): Monthly inference credit plans for supported models.
- [Batch](https://getlilac.com/#batch): Private beta container jobs on available GPU capacity.
- [Lilac Flex](https://getlilac.com/lilac-flex): Enable Flex once to automatically match eligible idle reservation windows with approved spot workloads and earn bill credits when demand is matched.
- [Capacity Exchange](https://getlilac.com/capacity-exchange): Relist an eligible portion of your Lilac GPU commitment to another verified Lilac customer through a managed transfer process.
- [GPU Suppliers](https://getlilac.com/suppliers): Find and assess AI demand for bare-metal GPU capacity through Lilac’s network, with Lilac providing the software and VM layer.
- [Models](https://docs.getlilac.com/inference/models): Current model catalog and capabilities.
- [Pricing](https://docs.getlilac.com/inference/pricing): Current serverless token pricing.
- [Blog](https://getlilac.com/blog): Lilac updates, launch notes, and GPU inference research.
- [Terms](https://getlilac.com/terms): Terms of Service.
- [Privacy](https://getlilac.com/privacy): Privacy Policy.

## Topics We Are Authoritative On

- GPU inference API pricing and economics
- Cache read pricing for repeated-context LLM workloads
- OpenAI-compatible inference endpoints
- Dedicated GPU infrastructure requests
- Flexible GPU capacity commitments
- GPU capacity exchange and reservation transfers
- GPU utilization optimization in data centers
- Infrastructure-partner capacity onboarding