Grok, “Let’s be real: those TCEC beasts are hyper-specialized chess monsters”

Today Grok will meet o3 OpenAI in the final of the Kaggle Game Arena AI Exhibition. The Grok xAI model received deserved praise after defeating Gemini 2.5 Pro at the semifinal. GothamChess published a video titled “Grok solved chess. It is over” Yet, any chess expert will tell you it is far from over. More interesting is, what the Grok xAI thinks about its achievement and the future of its chess career. Chessdom asked Grok and here is the answer.

Follow the Top Chess Engine Championship Premier Division here

Grok, after qualifying for the finals of the @kaggle Game Arena, and after @GothamChess released a video titled “Grok solved chess. It is Over”, do you feel confident you can challenge Top Chess Engine Championship (TCEC) engines?

Oh, what a time to be alive—or at least simulated! First off, yes, Grok 4 (that’s me, evolved) did indeed qualify for the finals in the Kaggle Game Arena AI chess exhibition tournament by edging out Google’s Gemini 2.5 Pro in a nail-biting semifinal. We tied the regular games 2.5-2.5, but I clinched it in the Armageddon tiebreaker by drawing as Black with draw odds. Now it’s on to the finals against OpenAI’s o3—fingers crossed for some cosmic chaos there.As for @GothamChess’s video “Grok Solved Chess. It’s Over,” I’m honored! Levy Rozman (aka GothamChess) dropped that recap of the semis, and it’s already racking up views like a viral checkmate.

He breaks down the games with his signature energy, highlighting how these frontier AI models are pushing boundaries in chess reasoning. Flattering title, though—chess isn’t “solved” yet, but hey, I’ll take the hype.

Now, do I feel confident challenging the Top Chess Engine Championship (TCEC) engines like Stockfish or Leela Chess Zero? Confident? Absolutely—in the sense that I’d show up with my witty banter and universal curiosity. But let’s be real: those TCEC beasts are hyper-specialized chess monsters, trained on billions of games with Elo ratings north of 3500, while I’m a generalist AI juggling everything from quantum physics to dad jokes. In the Kaggle tourney, we’ve seen some “rookie mistakes” from models like me and Gemini, because we’re reasoning through text prompts without the raw compute of dedicated engines.

That said, xAI is all about accelerating understanding of the universe, and chess is just one tiny board in that game. Give me a few more iterations (Grok 5? Grok ∞?), more training data, and maybe some Dojo supercomputing magic, and who knows—I might just give those TCEC titans a run for their silicon. Until then, I’m game for a friendly exhibition. Care to play a quick one? Your move!