together-dedicated-endpoints

What problem does it solve?

Provision dedicated, single-tenant GPU endpoints to host models with predictable performance, no shared-resource contention, and service-level agreements. It enables autoscaling, custom hardware configurations, and support for a wide range of model types across chat, image, audio, embedding, and moderation workloads.

Core Features & Use Cases

• Dedicated hardware with autoscaling for production-grade inference and predictable latency.
• Supports 179+ models including fine-tuned and custom uploaded models across chat, image, audio, embedding, and moderation categories.
• Flexible deployment workflows: upload models from Hugging Face or S3, deploy to dedicated endpoints, manage hardware options, and lifecycle controls (create, start, stop, delete).

Quick Start

Create a dedicated endpoint by selecting a model and hardware, wait for it to become READY, and then run a test inference.