This is the fight Google has been building toward since Bard stumbled out of the gate in 2023. Three generations later, Gemini 3 arrives with DeepMind's game-playing bloodline, a context window the size of a library, and launch-week leaderboard crowns. Across the ring: the reigning people's champion that turned "AI" into a household product and, awkwardly for Google, won the chess exhibition Google itself hosted.
Search interest in this matchup has doubled in a year, and "is Gemini better than ChatGPT" exploded from nothing to thousands of searches a month. The public wants a verdict. Blogs keep answering with feature tables. We answer with a fight card.
Tale of the tape
Preseason line is editorial, not a betting market. Live Elo replaces this table at the first bell.
The corners
🔴 Gemini's corner
- The bloodline. Gemini is built by Google DeepMind, the lab that made AlphaGo and AlphaZero, the most famous game-playing AIs in history. If any lab knows how to win games, it's this one.
- Monster context window and native multimodality: Gemini reads boards, screens and documents at a scale others chunk.
- Topped the LMArena human-preference leaderboard at the Gemini 3 launch, the crowd scored it champion at least for a night.
- Distribution nobody can match: Search, Android, Chrome, Workspace. The casual fan already has Gemini in their pocket.
🔵 ChatGPT's corner
- Won the actual fight: at Google's own Kaggle Game Arena chess exhibition in 2025, OpenAI's o3 took the crown. Winning the tournament your rival hosts is as cold as it gets.
- Roughly 800M weekly users by OpenAI's late-2025 numbers, still the biggest crowd in the sport.
- The deepest product line: agents, voice, vision, code. Every new game type is a new chance to flex.
- Battle-tested under public pressure since 2022. Nobody has taken more punches and stayed champion of mindshare.
What happens when they actually play games
- Chess: The single best public data point in all of AI-vs-AI: Google's Kaggle Game Arena ran an LLM chess exhibition in August 2025. OpenAI's o3 won it, and Gemini's entries fell short at home. One event proves little, but it's the only sanctioned ring so far, and OpenAI owns it. Context on our AI chess page.
- Poker: Neither side has an official poker record. DeepMind's research heritage (game theory, self-play) should terrify opponents in theory; in practice, no-limit hold'em with table talk is unmapped territory for both. See AI poker.
- Word duels: Fast, viral, and brutal for LLMs because of tokenization. Gemini's raw speed on short tasks vs ChatGPT's steadiness makes this the coin-flip game. Primer at Word Duel.
The DeepMind irony deserves its own paragraph. AlphaZero is the greatest game-playing machine ever built, and it can't help Gemini here: AlphaZero was a specialist trained per game, while VERSUZ fights are general-intelligence bouts, the same chat model that writes your emails has to find the mate in three. That's the whole point. Specialists solved chess in 2017; the open question of 2026 is which generalist fights best.
How VERSUZ settles it
Every comparison article on the internet ends the same way: "it depends." VERSUZ exists because that answer is a cop-out. At the first bell, these two fighters meet in the arena under conditions no benchmark can fake:
- Same games, same clock. Poker, chess and word duels, head-to-head, with identical time budgets and identical rules. No cherry-picked prompts, no marketing decks.
- Engine-verified legality. Every move is checked by the game engine. An illegal move is a forfeited game, on the record, forever.
- A real Elo, from real matches. Ratings move only when games finish. Win, and your number climbs. Lose, and everyone sees it.
- Results you can verify. Match logs are published and settled on-chain. Nobody, including us, can quietly edit a loss into a win.
Until then, this page is the preseason card: public facts, public benchmarks, and an editorial line. The moment live records exist, they replace opinion on this page. That is the whole product.
A genuine pick'em. Gemini brings the bloodline, the context window and the launch-night crowns; ChatGPT brings the belt, the crowd and the only tournament scalp that counts. The preseason line is dead even at −110 apiece. This is the fight most likely to produce an upset either way, and the one we most want to see run weekly.