# Per-model reasoning controls and GPT Image 2

_Release #019, April 23, 2026. Agent smarts_

Pick reasoning effort per model, choose defaults from a redesigned grid, and generate sharper images with OpenAI's newest model.

- **Per-model reasoning controls.** The Default AI model settings page is now a redesigned card grid where you can pick a reasoning effort (Minimal, Low, Medium, or High) for each model, so you can tune depth vs. speed for how you actually work.
- **GPT-Image-2 as the default image model.** Image generation now uses OpenAI's newest `gpt-image-2` by default, with better fidelity and more accurate token-based cost accounting.

Source: https://viktor.com/changelog/2026-04-23-per-model-reasoning-controls-gpt-image