---
canonical: "https://www.softatlas.io/together-ai/products/provisioned-throughput"
name: "Provisioned Throughput"
provider: "Together AI"
category: "Artificial Intelligence / AI Infrastructure / Inference Acceleration"
review_count: 0
ai_powered: true
website: "https://together.ai/provisioned-throughput"
updated: "2026-10-05"
---

# Provisioned Throughput

By [Together AI](/together-ai.md)

Token-based capacity with SLAs, providing committed inference capacity with reserved throughput and a 99% uptime SLA.

## Features
- Provide token-based capacity with SLAs
- Guarantee 99% uptime for inference workloads
- Offer reserved throughput for committed inference capacity
- Enable seamless migration from proprietary APIs
- Ensure predictable costs with PTU-based reserved capacity
- Guarantee hard performance with defined overage behavior
- Route overflow traffic to serverless fleet automatically
- Provide drop-in API compatibility

## Built for
DEVELOPERS, OPERATIONS, EXECUTIVES

Source: https://www.softatlas.io/together-ai/products/provisioned-throughput