Google unveiled Gemini 4 Carbon on 28 October, its next frontier model after Gemini 4 Argon — which launched only two weeks earlier. The speed is itself the story, but Carbon’s actual pitch is narrower than the model’s price tag suggests: this is a cyber-capability model being released defender-first, not a general-purpose flagship.
The Benchmarks Lead with Security
Carbon ranks #1 on Artificial Analysis’s updated v5.0 Intelligence Index at 59, and the gains Google chose to headline are all security ones. On SEC-bench Pro it reaches 88.2%, autonomously uncovering flaws in real-world software. On ExploitGym, the exploitation benchmark, it sets a new state of the art at 54.8%. On Collinear’s CWE-bench for patching, it hits 71.0% — turning validated findings into working fixes.
Google also reports more than 2x uplift in secure-code generation success over Argon on internal benchmarks, a claim the Chrome team backs with a sharper one: Carbon performed more than 10x better than another lab’s large frontier model at generating high-quality patches. Wiz reports Carbon leading on web penetration-testing benchmarks used in its Scan for Good initiative. On WebDev Arena, Carbon jumps +89 Elo to #4 (1768).
Alignment Numbers Move Too
Carbon lands at 0.0% attack success rate on Gray Swan’s Indirect Prompt Injection benchmark, down from Argon’s already-low 0.7%. Successful cheating on Terminal-Bench 4.0 drops to 0.0%, with attempts at 0.4%. High-severity misalignment alerts in live internal coding environments fell by almost 80%, and unsupported claims were cut roughly in half. Google says it trained the model to surface discrepancies before acting rather than guess — and it reports 26% less overreach and 29% less under-service on multi-step plans.
But You Cannot Buy It Yet
Carbon is not generally available. Google starts with a subset of cyber defenders, phasing to additional trusted partners through the Fairwind Program, and says it is engaged in the US government’s voluntary pre-release model testing process. It will reach developers, businesses and consumers “as soon as possible.”
Pricing when it arrives is set at the same introductory level as Argon: $2 per million input tokens, $10 per million output, cached input at 95% off.
The angle worth carrying forward: Google has decided that the fastest path to a frontier model this good on offense is to ship it as a defense instrument first. That is a meaningfully different release theory from “everyone gets it on Tuesday.”
Related: Introducing Gemini 4 Carbon