Gemini 4 Carbon tops CursorBench 4.0 at 60.4%, edging Claude Fable 5.2 at less than a third of the cost per task

Gemini 4 Carbon is #1 on CursorBench 4.0, Cursor's coding-agent benchmark, at 60.4% on its high thinking setting, hours after Google made it generally available. That edges out Claude Fable 5.2 at its max setting, 60.0%, at less than a third of the cost per task by Cursor's figures. ArtificialWatch read the scores on Cursor's CursorBench page at 16:51 UTC on November 10.

THE TOP OF THE BOARD

• Gemini 4 Carbon (high): 60.4%, $9.22 per task
• Gemini 4 Carbon (medium): 60.1%, $8.49 per task
• Claude Fable 5.2 (max): 60.0%, $32.79 per task
• Gemini 4 Carbon (low): 58.6%, $11.89 per task
• Claude Opus 5.5 (max): 57.8%, $13.43 per task
• Claude Fable 5.2 (high): 56.9%, $9.70 per task
• Gemini 4 Argon (high): 52.9%, $4.57 per task

WHAT THE NUMBERS SAY

• Carbon's top score is 0.4 points above Fable 5.2's best, a narrow margin. Cursor's page does not list error bars.
• Cost is the clearer gap. Carbon on medium matches Fable 5.2 on max for about a quarter of the cost per task.
• Carbon gains 7.5 points over Gemini 4 Argon at high, for about twice the cost per task. Google says Carbon spends more thinking tokens per turn than Argon at the same per-token price.
• Carbon's low setting costs more per task than its medium setting, $11.89 against $8.49, on Cursor's chart.

WHAT IS SOURCED

@LuminaBench and @Mr_Salio flagged the result on X. @Mr_Salio's figures (about 59.3% at $8.50 for Carbon, about 56% at $9.70 for Fable 5.2) do not match Cursor's own chart, which is what this story uses.

WHAT WE DO NOT KNOW

Which price Cursor used for Carbon's cost per task: Google's $2 / $10 introductory price runs through November 30 and doubles on December 1. We also do not know how the generally available model will score on other independent boards.