OpenAI GPT-6.2 Astra tops DesignArena game dev leaderboard; Sol ranks third
In brief
- GPT-6.2 Astra leads DesignArena's game dev leaderboard with a 1497 Elo rating as of November 21.
- GPT-6.2 Sol ranks third at 1430, 67 Elo points behind Astra.
- DesignArena rankings come from pairwise human preference votes, not a fixed exam.
- Both models reached the leaderboard roughly four days after the GPT-6.2 launch on or around November 17.
How DesignArena ranks models
DesignArena doesn't grade models on a fixed exam. It shows people outputs from two models side by side and asks them to pick the better one, and those pairwise human votes feed an Elo score. Each matchup comes down to one person's choice. The platform has evaluated more than 142 models this way, according to Crypto Briefing.
That's the key caveat for anyone reading the table. Rankings built from human preference votes show which outputs people picked in head-to-head matchups, so they aren't proof that one model is better across the board.
On the game dev board, the gap between Astra and Sol was 67 Elo points as of November 21.
Four days from launch to the top spot.
Who finished behind them
Astra and Sol finished ahead of several competitors, including Anthropic's Claude Opus 5.5 and Meta's Muse Wonder, according to the DesignArena standings reported by Crypto Briefing. The pair also showed up on a second leaderboard. Shortly after launch, Astra and Sol ranked #3 and #7, respectively, on Code Arena's WebDev leaderboard, Crypto Briefing reported. That's a lower placement for both models than they've posted on DesignArena's game dev board.
What OpenAI changed in GPT-6.2
Crypto Briefing said OpenAI's GPT-6.2 models improved on previous GPT-6 performance in areas such as visual reasoning and coding efficiency.
The report also carried two figures for Sol without naming where they came from. Sol reportedly delivers output quality comparable to earlier models at approximately 4x lower cost, and it's shown a 40% improvement in generation speed in certain evaluations, per Crypto Briefing. Neither number was tied to a named source in the report (so treat them as unverified claims, not benchmark results).
The DesignArena placements are the firmer data point here: they rest on human votes the platform tallies into Elo scores, and Astra's 1497 is the top mark on the game dev board right now.
Frequently asked questions
How does DesignArena rank AI models?
DesignArena doesn't grade models on a fixed exam. It shows people outputs from two models side by side and asks them to pick the better one. Those pairwise human votes feed an Elo score, and the platform has evaluated more than 142 models this way, according to Crypto Briefing.
What Elo scores did GPT-6.2 Astra and Sol get on DesignArena?
As of November 21, 2026, Astra held the top spot on DesignArena's game development leaderboard with an Elo rating of 1497. Sol placed third with 1430, a gap of 67 Elo points, Crypto Briefing reported.


