Elon Musk released a new language model and, as always, graded it himself. xAI, Musk's company, unveiled Grok 4.5 on Wednesday, and its owner immediately described it as "a model in the Opus class, only faster, more token-efficient and cheaper." Internally, Musk claims, the model was "roughly comparable to Opus 4.7, but much faster." Notice how the sentence goes: not "the best in the world," but "comparable to the best, only cheaper." That's a salesman's positioning, not an engineer's assessment.

The numbers, though, say something concrete. Grok 4.5 costs 2 dollars per million input tokens and 6 dollars per million output. For comparison, Anthropic's Claude Opus 4.7 charges 5 dollars for input and a full 25 for output - nearly four times more on the output side. OpenAI's most expensive model goes up to 30 dollars per million output tokens. If price is the only criterion, Musk really does have something to show.

But "cheaper" doesn't mean "better." According to the available tests, Grok 4.5 is competitive but, as analysts note, "a little below the top." That's an important distinction in a world where every new model is pitched as a leap into the future. This one isn't a leap - it's a step sideways: less money for nearly the same work. For companies spending millions on artificial intelligence every month, that very maths is worth more than any marketing superlative.

The context is worth mentioning too. The model comes soon after Musk's company went public, and nothing drives valuation like the claim that your product is neck and neck with the most expensive on the market, but at half the price. Whether Grok 4.5 really is "in the Opus class" will be shown by the independent tests in the coming weeks - not by a statement from the man for whom that statement directly lifts the share price.