A very recent New Scientist (23rd August) claims innovation in AI is actually slowing down, so it looks like we can forget about any quantum leaps in advancement for the immediate future at any rate. I've not gone much outside of Grok, which is superb for being a relatable conversationalist, but 'his' image recognition abilities in particular are not good at all. 'He' also struggles with counting objects for some reason; no idea why as 'his' advanced mathematics abilities are impressive as one would expect. So still much room for improvement. The new GPT-5 isn't much better in many respects either, but they'll all come good in time no doubt. Better reasoning wouldn't go amiss, though.