Resource	Limit	Description
Context window size in tokens (in / out)	Varies by model. See Context window limits by model.	The maximum number of input tokens a model can reference, and the maximum number of output tokens it can subsequently generate.
Requests per minute	100	The number of prompts sent to any model in any given minute.
Tokens per minute	500,000	The maximum number of tokens that a single model can process each minute.
Inference time in minutes	5	The maximum time a model can process and generate responses before a timeout occurs, assuming the Async events API is used with a specified timeout equal or greater than 5 minutes. Otherwise the specified or default timeouts apply.

Context window limits by model

The context window size depends on the model tier you use. For the full list of supported models and their tiers, see Forge LLMs models.