The most interesting part, even more than ARC 3 score, to me is that this is the first model I recall seeing that scores lower on Max than High reasoning effort on some coding benchmarks:

Terminal-Bench 4.0: High (57.9%), Max (56.7%)

DeepSWE: High (73.3%), Max (71.5%)

It _loses_ 1-2% performance going to High from Max

▲

XCSme 5 hours ago | parent [-]

That's quite common with many models, after "High" reasoning, over-thinking starts occurring and the model skips over the right solution by convincing itself otherwise.

▲

m0zzie an hour ago | parent | next [-]

I find this very amusing, given we humans are also highly susceptible to this.

▲

GodelNumbering 5 hours ago | parent | prev [-]

> That's quite common with many models

Such as?

I can't think of any. Diminishing returns, yes. Occasionally flat, yes. Downright regression, no.

	▲	XCSme 5 hours ago \| parent \| next [-]
		In my own tests on aibenchy.com, where questions are quite simple, higher reasoning efforts consistently used to do worse than medium for most models. The reasoning effort should match the complexity of the task against the model's capability. Hard task with low reasoning = bad Easy task with very high reasoning = bad
	▲	minatoaqua1 4 hours ago \| parent \| prev [-]
		grok 4.6