Google announces Gemini 4 Carbon, its next frontier model, to cyber defenders first

Google announced Gemini 4 Carbon on October 28, 2026, calling it its strongest model, with state-of-the-art cybersecurity and long-horizon capabilities. It follows Gemini 4 Argon by two weeks. Sundar Pichai's post and Google's blog say it is not yet generally available: starting this week a subset of cyber defenders get access, with more trusted partners added through the Fairwind Program. Google says it will bring Carbon to developers, businesses and consumers "as soon as possible", and that it is taking part in the U.S. government's voluntary process for pre-release model testing.

Price (Google): the same introductory price as Argon, $2 per million input tokens and $10 per million output tokens, with cached input 95% cheaper.

Google's claims (its own post, not independently verified):

• Carbon ranks #1 on Artificial Analysis Intelligence Index v5.0 at 59.
• Cybersecurity: 88.2% on SEC-bench Pro (finding vulnerabilities), 54.8% on ExploitGym (exploitation) and 71.0% on CWE-bench (patching), evaluated without guardrails. Google says it generates secure code at more than 2x Argon's success rate, and that the Chrome team found its patches more than 10x better than another lab's large model.
• Coding: it says Carbon jumps 89 Elo to #4 (1769) on WebDev Arena, and has run for days on multi-step engineering tasks inside Google.
• Safety: successful cheating on Terminal-Bench 4.0 reduced to 0.0% (attempts 0.4%), high-severity misalignment alerts in live coding down almost 80%, unsupported claims cut about 2x versus Argon, overreach down 26% and under-service down 29%, and a 0.0% attack success rate on Gray Swan's indirect prompt-injection benchmark.

Artificial Analysis (a benchmark site, its own measurements) scores Carbon (high reasoning) 59 on its Intelligence Index v5, #1 on 4 of the 9 evaluations (HLE-Diamond, Terminal-Bench-Science, AA-Omniscience, AutomationBench-AA). In Antigravity CLI it scores 59 on its Coding Agent Index, behind only Claude Code with Claude Opus 5.5 (62) and 6 points above Argon (53). It reports a cost of $7.76 per Intelligence Index task, about 23% less than Claude Opus 5.5 (max). On AA-Omniscience Carbon scores 54, with accuracy up to 58% and a hallucination rate up from 4% to 11%. It lists a 1M-token context window, text, image, video and speech input, and standard pricing of $4/$20 per million tokens with Google's 50% launch discount, which has no confirmed end date.

Arena (a leaderboard publisher, its own measurements) says Carbon (High) is #1 in Text Arena with 1528 points, #2 in Agent Arena at a median $1.80 per task (72% less than #1 Claude Opus 5.5 Max), and #4 in Code Arena: WebDev with 1768 points.