Gemini 4 Carbon ranks second with 60 points

Google's Gemini 4 Carbon ranks second on the Artificial Analysis Intelligence Index with 60 points, the test site showed on the 14th of November 2026. Only Anthropic's Claude Fable 5.2 scores higher, with 65.

The index is an outside ranking of nine hard tests. Google opened Carbon to all paying developers on the 10th of November, the company said.

Second on two rankings

Carbon comes ahead of Anthropic's Claude Opus 5.5 at 59. On MKA, a new test of office work with drawings and scans, Carbon scored 54 percent, Artificial Analysis said.

Fable 5.2 scored 57 percent and Opus 5.5 scored 53 percent.

First place in maths

On MathArena, a maths ranking from ETH Zurich, Carbon took first place on the 13th of November. It scored 85.4 percent to 82.8 percent for Fable 5.2.

What it is for

Google said Carbon is built for long coding jobs, science research and finding security flaws. A swarm tool in Google Antigravity lets one request start a team of AI helpers.

Price

Developers pay 2 dollars per million input tokens and 10 dollars per million output tokens, Google said. After the 30th of November that doubles to 4 and 20 dollars.

What went viral on X

The MathArena result post by Jasper Dekoninck had about 107,000 views.

Below we break down the cost

On MKA, a Fable 5.2 task cost about 31.97 dollars and a Carbon task about 6.29 dollars, Artificial Analysis said. Fable costs about 5 times as much for 3 more points, by our arithmetic.

What is confirmed and what is not

Google helped design MKA, and an OpenAI model grades it, Artificial Analysis said.

Key facts

Google opened Gemini 4 Carbon to all paying developers on the 10th of November 2026.

Carbon scores 60 on the Artificial Analysis index, behind Fable 5.2 at 65.

It ranks second on the MKA office test and first on MathArena.

Its starting price doubles after the 30th of November.