A comparison published in the period showed equivalent performance between models from two frontier vendors on threat detection tests.
The result bears directly on the regulatory debate that would dominate the following weeks. If the capability that worries is distributed across vendors, restricting one displaces usage rather than reducing risk.
Security leaders would make that argument in their petition against models being pulled: anyone wanting offensive capability finds an alternative, while defenders lose the tool they were using within the rules.
There's a second, less discussed effect. A tie on capability benchmarks means the difference between vendors shifts to price, availability and contract terms, which is exactly where competition ended up over the semester.
For anyone choosing a model, the data suggests generic capability comparisons have limited value. What decides is measuring on the operation's specific task, with real data, comparing cost per useful result.