About Salad
Salad runs containers and AI models across a distributed network of GPUs. It aims at teams that want lower GPU costs than major hyperscale clouds.
What it does
You deploy a container group, choose a GPU class and set the number of replicas, and the platform schedules the work across available nodes. A managed token API offers open-weight models through an OpenAI-compatible endpoint. Billing starts only when each instance is running, which keeps idle time out of your bill.
Key features
- Container deployments across many GPUs
- Managed token API for open-weight models
- Per-second billing for running instances
- Community and reserved GPU pools
- Private deployment option for enterprises
Who it's for
- Teams with batch or background AI jobs
- Companies looking for lower GPU prices
- Developers who want an API for open models
Pricing and good to know
GPU prices start from a low hourly rate and vary by GPU class. Reserved capacity and private deployments are arranged for enterprise customers. Because capacity is spread across a distributed network, plan your jobs so they can restart cleanly if an instance is reallocated.
Highlights
- Deploy container groups across many GPUs
- Call a managed token API for open models
- Pay only while instances are running
- Choose community or reserved GPU pools
- Run private deployments for enterprises
