Provider

The same model can be available in many places, with different prices, speed, limits and data rules. Picking a provider means picking those conditions.

The four kinds

• Labs: train the models and sell direct access through their own API.
• Clouds: resell models from many labs inside the cloud account a company already uses. Examples as of November 2026: Amazon Bedrock, Google Gemini Enterprise Agent Platform (formerly Vertex AI) and Microsoft Foundry.
• Inference providers: run open-weight models with a focus on cost and speed, such as Together AI, Fireworks, Groq and Cerebras.
• Routers: give you one API and one bill for hundreds of models across providers, such as OpenRouter. They don’t run the models, they route requests.

What to check

Where your data lives, whether it is used for training, price per million tokens, rate limits, and whether your company’s contract already covers that provider.