Open, convenient and predictable: Introducing Provisioned Throughput

Introduces Provisioned Throughput (PTU) as a new inference form factor for open-weight models. PTUs offer guaranteed capacity with a 99% uptime SLA and token-based pricing, distinct from serverless (best-effort) and dedicated inference (full control). The pricing model accounts for input, cached input, and output tokens at different rates, enabling cost optimization. The post highlights the economic benefits compared to proprietary models and positions PTUs for production workloads requiring reliability and predictable costs.