Google's Gemini 4 Carbon takes #1 on Artificial Analysis' Intelligence Index

Key figure: 59, Gemini 4 Carbon’s score on the Artificial Analysis Intelligence Index v5, the top score on the index. Source: Artificial Analysis.[2]

What happened

Google announced Gemini 4 Carbon on Oct. 28, calling it “our next frontier model.” Chief executive Sundar Pichai shared it on X at 1:01 p.m. ET, two weeks after Google opened Gemini 4 Argon to paying users.[4][5] Google says Carbon is its strongest model, with “state-of-the-art cybersecurity and long horizon capabilities.”[1]

The ranking

Google says Carbon ranks #1 on version 5.0 of the Artificial Analysis Intelligence Index, at 59.[1] Artificial Analysis, an independent benchmarking firm, lists it first on its own leaderboard, one point ahead of Anthropic’s Claude Opus 5.5 at its highest setting. Argon scores 50.[2]

Exhibit 1. Artificial Analysis Intelligence Index v5, top scores

Model (setting)	Developer	Score
Gemini 4 Carbon (high)	Google	59
Claude Opus 5.5 (max)	Anthropic	58
Claude Opus 5.5 (xhigh)	Anthropic	57
GPT-6 Astra (max)	OpenAI	57
GPT-6 Astra (xhigh)	OpenAI	55
Gemini 4 Argon (high)	Google	50

Source: Artificial Analysis leaderboard, checked 1:35 p.m. ET, Oct. 28, 2026. Table: New Street.

Artificial Analysis also lists limits. On its Coding Agent Index, Carbon scores 59, second to Claude Opus 5.5 at 62. Carbon answers more questions correctly than Argon on the firm’s knowledge test, but its hallucination rate rose to 11% from Argon’s 4%, which the firm says is still the second lowest of any model it has tested. It also says Carbon is not leading in coding and knowledge work overall.[3]

What Google reports

Google reports that Carbon scores 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching, measured without guardrails. The company also reports a #4 rank on WebDev Arena at 1769, up 89 points, and more than twice Argon’s success rate at generating secure code in internal tests. These are Google’s own figures.[1]

On safety, Google says Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0% and cut unsupported claims by about half versus Argon in its internal testing.[1]

What's known, and what isn't yet

Access is staged. Starting this week, Google is giving Carbon to a subset of cyber defenders, then to more trusted partners through its Fairwind Program. It says developers, businesses and consumers will get it “as soon as possible,” with no date. Google also says it is taking part in the U.S. government’s voluntary pre-release testing process.[1]

Pricing is set but not yet live. Google says Carbon will launch at the same introductory price as Argon, $2 per million input tokens and $10 per million output tokens, with cached input 95% off, then $4 and $20.[1] Google has not given an end date for Carbon’s introductory period. As of Oct. 28, Google’s API pricing page lists Gemini 4 Argon but no Carbon model ID.[6]

What comes next

The next marker is wider access: the date Carbon reaches the Gemini API and Google’s apps, and the price developers pay when it does. Until then, the ranking gives a clear read on where Google stands among the top models.