Musk says Grok V9 is trained. We've heard this tune.

On 25 May, Musk said V9-Medium — reportedly ~1.5T parameters, ~3× the current production model — had finished training with 'positive' evals. A mid-June release got floated. By mid-June: no firm date, no model card, no independent numbers. Meanwhile the much-hyped Grok 5 is still training and the betting markets don't believe a June ship.

This isn't generic cynicism — it's pattern-matching. xAI's timelines have slipped repeatedly, 'positive internal evals' is the least falsifiable phrase in AI, and parameter count (1.5 trillion!) is the vanity metric that says nothing about whether the thing is actually good. The shipping flagship is still Grok 4.3. Until V9 is something you can call, it's a press cycle.

So: noted, not believed-yet. The grown-up move is to ignore the announcement and wait for the artefact — model card, third-party benchmarks, an API you can hit. When those exist, we'll cover the model. Right now we're covering a claim.