AI models overview
Which AI models Coinrule uses internally, their credit multipliers, and how model selection works.
Coinrule uses Anthropic Claude models internally to generate and test trading strategies — currently claude-haiku-4-5 for strategy parsing and claude-sonnet-4-6 for background test generation. Model selection is automatic; users do not choose which model is used.
Which models are in use?
| Model ID | Credit multiplier | Usage in Coinrule |
|---|---|---|
| claude-haiku-4-5 | 1× | Strategy parsing (main) |
| claude-sonnet-4-6 | 3× | Test generation (background) |
| claude-opus-4-5 | 15× | Reserved / not user-selectable |
How does model selection work?
Strategy parsing always uses claude-haiku-4-5. This model handles the primary parse of your strategy description into executable Python. It is fast and cost-efficient (1× multiplier).
Test generation runs in the background after you launch a strategy and uses claude-sonnet-4-6. Tests are required before the engine can load the strategy, but this step happens automatically — you don't need to do anything.
claude-opus-4-5 is defined in the credit pricebook for potential future use but is not currently used in any user-facing flow.
Can I bring my own API keys?
Custom model configuration (bring-your-own API key, model overrides) is not currently available to users. The Models & APIs tab in Settings is where this feature will appear when it ships.
Credit cost
See Credit drivers and costs for the LLM credit formula.