AI Models behind Minds

Different parts of the platform use different AI models — picked per task for the best trade-off between quality, speed, and cost. The table below is grouped by functional area and fetched live from the running production prompts, so it always reflects what the platform is actually using right now.

How to read this

• Area is a functional grouping inside the platform (Mind chats, image analysis, panels, voice, …).
• Provider is who hosts the model — Anthropic (Claude), OpenAI (GPT / o-series), or Google (Gemini).
• Model is the underlying AI model name (e.g. claude-sonnet-4-6, gpt-4.1-mini, gemini-2.5-flash). If an area runs more than one task internally, the deduped list of all models used is shown.

Overriding models per request

If you're calling the platform programmatically, you can override the model on a per-request basis (for benchmarking, cost tuning, or testing new releases) by passing model and provider parameters in your API call.

See the API model override docs for the full list of supported providers and the auto-detection rules.

Keeping up with releases

Default model selections are reviewed regularly and updated whenever providers ship a meaningful upgrade. The table above will reflect the change the moment the new defaults are deployed — no manual doc edits needed.