# Claude Haiku 4.5 Is Now the Default AI Model for All New Tests

> "We switched the default AI agent from Claude Sonnet to Claude Haiku 4.5. Tests run faster. Existing tests

Source: https://qa.tech/changelog/claude-haiku-4-5-default-ai-model · Published: 2025-12-11

---
We switched the default AI agent from Claude Sonnet to Claude Haiku 4.5. Tests run faster. Existing tests are unaffected.

**What Changed**

-   **New default:** New tests use Claude Haiku 4.5 unless you select a different agent
-   **Existing tests:** Keep their current agent assignment. No action required
-   **Switching agents:** Open test settings or expand Advanced during test creation to select Claude Sonnet 4.5 if needed

Haiku 4.5 handles most test scenarios. Use Sonnet 4.5 for tests that require more complex reasoning across multiple steps. See [agent selection](https://docs.qa.tech/best-practices/creating-tests#choosing-an-ai-agent) for comparison.