Breaking

Google's Unreleased Gemini 4 Carbon Reportedly Matches Anthropic's Opus 5.5 on Artificial Analysis Index

November 3, 2026 at 13:15 EST

Foundation Models

Google's unreleased Gemini 4 Carbon model, run at its MEDIUM reasoning effort setting, reportedly scores 58 on the Artificial Analysis (AA) index, level with Anthropic's Opus 5.5 at its maximum effort setting and one point ahead of OpenAI's GPT-6 Astra, which scores 57.

Reported · Artificial Analysis index

Gemini 4 Carbon reportedly matches Opus 5.5 while running at a lower effort setting

An unreleased Google model scores 58 on the AA index at medium effort, level with Anthropic's Opus 5.5 at maximum effort and one point above OpenAI's GPT-6 Astra. The figures are unverified.

Gemini 4 Carbon

Medium reasoning effort

Opus 5.5

Maximum effort

GPT-6 Astra

Reported score

AA index score, drawn to scale

Column height is proportional to score, starting from zero. Gridlines every 10 points.

6050403020100

Gemini 4 Carbon
Medium

Opus 5.5
Maximum

GPT-6 Astra
Reported

The gap between first and third is one point, which is small enough that a change in test sets or settings could reverse it.

The central caveat

Matching scores at different effort levels is not a like-for-like comparison.

• Gemini 4 Carbon is reported at medium effort,

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.