Open, convenient and predictable: Introducing Provisioned Throughput

2026-07-08 · Source: Together AI | The AI Native Cloud - Together.ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Together AI has introduced Provisioned Throughput, a new inference offering providing reserved capacity for frontier open models with token-based pricing and a 99% uptime SLA. This service addresses the gap between best-effort serverless and complex dedicated inference, offering the simplicity of pre-optimized models with guaranteed capacity. It is available today for MiniMax M3 and GLM-5.2, with capacity in North America and EMEA, requiring a one-month minimum term. Provisioned Throughput aims to significantly reduce inference costs, claiming up to 90% below Claude Opus 4.8 list price. Companies using open models have reported 6-20x lower inference costs compared to proprietary alternatives, with Together AI's API token volume growing from 30 billion to over 400 trillion tokens monthly in nine months. This solution allows companies to scale multi-model, multi-harness setups for agents in coding, finance, and marketing.

Key takeaway

For AI Engineers or MLOps teams managing production workloads with open models, Provisioned Throughput offers a predictable, cost-effective alternative to best-effort serverless or complex dedicated GPU setups. If you are currently using proprietary APIs for tasks suitable for frontier open models, you can achieve up to 90% cost reduction with guaranteed capacity and a 99% uptime SLA. Consider migrating your existing workloads to MiniMax M3 or GLM-5.2 to benefit from predictable token-based pricing and simplified infrastructure management.

Key insights

Provisioned Throughput offers guaranteed, token-based inference capacity for open models, bridging the gap between serverless and dedicated GPU options.

Principles

• Abundant AI needs committed capacity.
• Match task difficulty with model quality.
• Open models offer significant cost savings.

Method

Customers purchase Provisioned Throughput Units (PTUs), which are fixed slices of guaranteed token capacity per minute for a specific model, priced at \$0.05 per PTU per minute. PTUs are consumed differently by input, cached input, and output tokens.

In practice

• Migrate production workloads from proprietary APIs.
• Optimize spend by understanding PTU consumption rates.
• Utilize MiniMax M3 or GLM-5.2 for guaranteed capacity.

Topics

• Provisioned Throughput
• Open Models
• AI Inference
• Token-based Pricing
• SLA
• MiniMax M3
• GLM-5.2

Best for: CTO, VP of Engineering/Data, Entrepreneur, MLOps Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• not much happened today — Rapid advancements in non-invasive brain-to-text, efficient AI inference, and sophisticated agent orchestration are reshaping AI development.
• A practical guide for platform teams managing shared AI deployments — Rate limiting protects model health; quota reservations ensure critical workload priority in shared AI deployments.
• OpenRouter Now Processes More Than a Quadrillion Tokens a Year — OpenRouter provides a unified API for 400+ AI models, optimizing access, routing, and cost-quality balance for diverse tasks.
• Top Cost-Efficient Small Models for AI APIs — Small language models offer significant cost, latency, and privacy advantages for API builders through efficiency gains.
• OpenAI Discovers New Way to Cut Inference Costs in Half — OpenAI significantly cut inference costs by optimizing existing server chips.
• not much happened today — AI development is shifting towards multi-cloud distribution, cost-aware agentic workflows, and specialized hardware for inference.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Together AI | The AI Native Cloud - Together.ai.