Choose a model
Choose an AI model for a Jotsu application by matching capability, context, latency, and cost to the specific task users need to complete.
Choose a model for a specific AI step, not for the application as a whole unless every step has similar needs.
Use a higher-capability model for ambiguous reasoning, coding, tool selection, or high-value synthesis. Use a smaller model for classification, extraction, routing, and repetitive transformations with a clear schema.
Compare capability, latency, price, context size, tool use, and structured-output reliability. Set a token limit, timeout, and bounded retry policy.
Model selection is currently configured in application code. Ask Codex to identify the existing inference call and change only its model configuration. Test fixed examples before and after the change.