Does setting the /effort to xhigh or max actaually change anything with free only models?

[giuseppeeeeeeeeee]

Does the /effort actually change anything or is it just that Claude Code thinks its going xhigh or max?

[Samielakkad]

It is not only Claude Code ?thinking? the effort changed: 9Router does read the requested effort and translates it for the selected provider. However, the upstream model still decides what levels are actually supported.

For the free providers, the important part is that the levels often collapse:

• Kiro: xhigh and max are both normalized to high.
• Gemini/Vertex: unsupported xhigh or max is clamped to high.
• OpenAI-format models: max is clamped to xhigh, because that format has no max value.
• A model/provider without configurable reasoning may have the parameter removed or ignored.

So low ? high can make a real difference on a reasoning-capable free model, but xhigh ? max will usually make no additional difference for those free routes. It cannot give a model more reasoning capability than its upstream API exposes.