Google's Gemini 4 Carbon has taken the top spot on the MathArena math benchmark, the first non-OpenAI model in a while to do so, according to The Decoder. It displaced GPT-6.1 Sol, and Claude Fable 5.2 placed second.

The report says Carbon may be one of the most capable publicly available models for math tasks. OpenAI's models, which the report says recently solved demanding equations such as the Navier-Stokes equations, are not publicly available yet.

Carbon's most common failure mode is exceeding the time limit. The report also says the model almost never gives up on a task, a trait the report notes 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.

The higher performance comes at a cost. Using the reasoning settings from testing, the report says Carbon and Fable 5.2 cost roughly 15 and 40 times as much as GPT-6.1 Sol, respectively.