ENDE
AI

Google unveils Gemini 4 Carbon, a cybersecurity-focused frontier model for defenders first

Announced Oct. 28, 2026, Carbon tops Artificial Analysis' Intelligence Index v5.0 at 59, Google says, and goes first to a subset of cyber defenders via the Fairwind Program.

By ·

Google unveils Gemini 4 Carbon, a cybersecurity-focused frontier model for defenders first
AI-generated illustration

TL;DR

  • Google announced Gemini 4 Carbon on Oct. 28, 2026, calling it its strongest model, with a focus on cybersecurity and long-horizon tasks.
  • Google says Carbon ranks No. 1 on the Artificial Analysis Intelligence Index v5.0 at 59 and scores 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench.
  • Access starts with a subset of cyber defenders, then trusted partners through Google's Fairwind Program; no date is given for general availability.
  • Introductory pricing is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 afterward.

Google's newest frontier model is built to find and fix bugs, and the public can't use it yet. On Oct. 28, 2026, Google announced Gemini 4 Carbon, "our strongest model with state-of-the-art cybersecurity and long horizon capabilities," according to Google DeepMind's Koray Kavukcuoglu.

Gemini 4 Carbon security benchmark scores, per Google
SEC-bench Pro (find): 88.2%; CWE-bench (patch): 71.0%; ExploitGym (exploit): 54.8%SEC-bench Pro (find)88.2%CWE-bench (patch)71.0%ExploitGym (exploit)54.8%
Source: Google

The numbers Google is claiming

  • 59 on the Artificial Analysis Intelligence Index v5.0, ranked No. 1. The Verge reports Claude Opus 5.5 and GPT-6 Astra trail behind.
  • 88.2% on SEC-bench Pro (finding vulnerabilities), 54.8% on ExploitGym (exploitation) and 71.0% on CWE-bench (patching).
  • More than 2x uplift over Gemini 4 Argon in generating secure code on internal benchmarks; +89 Elo to No. 4 (1768) on WebDev Arena.
  • Successful cheating on Terminal-Bench 4.0 cut to 0.0% and a 0.0% attack success rate on Gray Swan's Indirect Prompt Injection benchmark, per Google.

Defenders first

Google says it is "actively engaged in the U.S. government's voluntary process for pre-release model testing." Carbon goes first to a subset of cyber defenders, then gradually to trusted partners through the Fairwind Program, before reaching developers, businesses and consumers. The Verge notes Carbon follows Gemini 4 Argon, released earlier in October 2026.

Pricing

Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens, with cached input 95% off. After the introductory period, Google lists $4 and $20.

What it means for Web3 teams – Analysis

Models that find and exploit real vulnerabilities cut both ways for smart-contract projects: defenders get faster audits, attackers get cheaper bug hunting. Benchmarks are Google's own claims. Expect AI-assisted code review to become a baseline expectation before mainnet launches. This is not investment advice.

FAQ

What is Gemini 4 Carbon?

Google's frontier model announced Oct. 28, 2026, focused on cybersecurity and long-horizon coding and agentic tasks.

Can developers use Gemini 4 Carbon now?

Not yet. Google is giving access first to a subset of cyber defenders and then trusted partners via its Fairwind Program, before a wider release.

How much will Gemini 4 Carbon cost?

$2 per million input tokens and $10 per million output tokens at introductory pricing, then $4 and $20.

How does it score on security benchmarks?

Google reports 88.2% on SEC-bench Pro, 54.8% on ExploitGym and 71.0% on CWE-bench.

Sources

  1. Google – Gemini 4 Carbon: our next frontier model
  2. The Verge – Google launches Gemini 4 Carbon

News summary based on the sources above. Not investment, legal or tax advice. Editorial policy & corrections · Report an error

Building in Web3? Turn the news into trust.

Positioning, KOL programs, LinkedIn systems and claim-safe content for Web3 and fintech teams.

Book a free 30-min callSee services

Related

Book a free call