Elon Musk Shelves Grok 4.7 and Unveils Grok 4.8 Instead
Article excerpt
Highlighted: the sentence this signal was extracted from
Grok 4.7 still isn't out, and that is the real story. Elon Musk has explained the delay as a reinforcement learning problem, but xAI has not published a model card, price, benchmark, or release date for the model developers are waiting to test. If you've been waiting on xAI's next Grok, here's where things actually stand. On September 11, Musk wrote on X that Grok 4.7 needed "a few more days to cook." He said the team may have penalized response length too much during reinforcement learning. That left the model too quick to give up on hard tasks it can solve, and too sloppy about checking its own work. That part matters. It isn't a cosmetic bug or a launch-page typo. The flaw Musk described goes straight to the sort of work xAI wants Grok to handle: coding, math, research, agents, and long tasks where the final check is often the difference between useful output and expensive nonsense. xAI has still not put Grok 4.7 into the market. As of September 14, the company's own news page lists Grok 4.6 as the latest released flagship model. There's an August 12 launch note, API access, and availability on Cursor and Grok Build, with pricing at $2 per million input tokens and $6 per million output tokens. There is no official Grok 4.7 product page, no API identifier, no public price - and no benchmark table either. You don't need to be an AI researcher to see the problem. Musk has...
Keep reading with a free account
The rest of this article, and every signal for xAI, is in your free account.
