Gemini 4 Carbon Tops New AI Index v5, Edging Claude Opus 5.5 by One Point

Artificial Analysis published the Intelligence Index v5 on October 27 and then added Gemini 4 Carbon's result on October 28, placing the Google model first with 59 points, according to aitimes.com. Claude Opus 5.5, which the evaluation company had initially presented as the leader on October 27, finished second with 58 points.

GPT-6 Astra took third place with 57 points, followed by GPT-6.1 Sol with 55 points in fourth and Claude Sonnet 5.5 with 54 points in fifth. Gemini 4 Carbon recorded the best score in four of the index's nine evaluations: difficult knowledge and reasoning, scientific research, knowledge accuracy and task automation.

New benchmarks and more private data

The v5 index weights knowledge work, coding, science and general knowledge at 25% each. Artificial Analysis introduced two new benchmarks, FrontierCode 2.0 and Terminal-Bench-Science, and improved three existing evaluations. On Terminal-Bench-Science, GPT-6 Astra scored 63%; it scored 72% on CritPt v2.

The share of private data in the overall evaluation rose from 45% to 60%, and the median number of output tokens per evaluation task increased from 35,000 to about 59,000. Kimi K3.1 Pro was the highest-scoring open-weight model with 46 points, 13 points behind Gemini 4 Carbon.

Artificial Analysis also launched a Progress Index that compares model performance on a single scale going back to November 2022. In January 2026, the company had released version v4 of the Intelligence Index.