Grok 4.5 is in private beta, and every performance claim is unverifiable

Elon Musk said on June 28 that xAI’s Grok 4.5 is running in private beta with teams at SpaceX and Tesla, built on a new 1.5-trillion-parameter foundation, with internal evaluations he says put it near or past Claude Opus. Here is what you can independently check against that claim: nothing.

Nobody outside those two companies has access to the model. It has not been submitted to a single public benchmark. The “near or past Opus” line is the CEO describing his own team’s private eval, which is a press release wearing a lab coat. It might even be true. You have no way to know, and neither do I, and that is exactly the problem.

lvl30 does not cover a model as a release until there is something to run, or an API to hit, and a number produced by someone other than the vendor. A closed beta at the founder’s other two companies is not that. It is a teaser with a parameter count attached.

When Grok 4.5 actually opens up, it goes through the same gauntlet every frontier model here runs, on real work, scored on what it does instead of what its maker says it does. Until then the honest headline is the one at the top: it exists, it is closed, and the benchmarks are a rumor.