Gemini 4 Carbon tops MathArena, dethroning GPT-6.1 Sol

The MathArena leaderboard had been led by OpenAI models, so Gemini 4 Carbon's rise to the top marks a shift in which company holds the benchmark crown. OpenAI's models, which recently allegedly solved demanding equations like the Navier-Stokes equations, aren't publicly available yet, meaning the comparison is among models users can actually access.

Gemini 4 Carbon's most common failure mode is exceeding the time limit, and the model almost never gives up on a task—a trait that can be read as either positive or negative. In one case, it launched a script that ran for more than eleven hours without checking on its progress. That persistence likely contributes to both its benchmark score and its cost profile in testing.

The cost gap is pronounced: with the reasoning settings used in testing, Carbon and Fable 5.2 cost roughly 15 and 40 times as much as GPT-6.1 Sol, respectively. For teams weighing benchmark performance against budget, the leaderboard win may come with a steep trade-off.