Add a New Model to the Eval Runner

Follow these steps whenever the user asks to add a new AI model to the eval suite.

Step 0: Gather Information

Determine the following (ask the user if not provided):

1. Model identifier - the OpenRouter-style name, e.g. anthropic/claude-opus-4.6. If the user gives a marketing name or URL, look up the OpenRouter model id.
2. Formatted name - human-readable, e.g. Claude 4.6 Opus.
3. Provider family & version - needed to find older siblings (e.g. claude-opus-4.5 is the predecessor of claude-opus-4.6).
4. apiKind - only needed for OpenAI Codex/Responses-API models; set to "responses". Omit for all other models.

If you're unsure, check how the closest existing model in the same family is configured in runner/models/index.ts and match it.

Step 1: Add the Model to runner/models/index.ts

Open runner/models/index.ts and add a new entry to the ALL_MODELS array. Place it next to its family siblings, respecting the existing grouping comments.