Back home

Gemini 4 Carbon

x 0.90
Visit source

Intro

Gemini 4 Carbon is the frontier model Google DeepMind announced on October 28, 2026, just two weeks after Gemini 4 Argon. Google calls it "our strongest model with state-of-the-art cybersecurity and long horizon capabilities". The announcement was posted by Koray Kavukcuoglu (SVP, Google DeepMind and Chief AI Architect) and was covered the same day by Reuters, Ars Technica and 9to5Google; "Gemini 4 Carbon" also became a trending topic on X in the US.

Official post: Introducing Gemini 4 Carbon · Benchmark write-up: Artificial Analysis

Explain

Key points

  1. Top of the leaderboard: Carbon scores 59 on the Artificial Analysis Intelligence Index v5.0, ranking #1 ahead of Claude Opus 5.5 (58) and GPT-6 Astra (57). On the Artificial Analysis Coding Agent Index it is second only to Claude Opus 5.5.
  2. Cybersecurity focus: state-of-the-art results across the defense lifecycle: SEC-bench Pro 88.2% (finding vulnerabilities), ExploitGym 54.8% (exploitation) and CWE-bench 71.0% (patching). On Google's internal benchmarks it is more than 2x as successful as Argon at generating secure code, and the Chrome team reports much better patch generation.
  3. Coding and agents: Google says it improved coding, agentic and scientific tasks, full-stack web development and multimodal understanding. On WebDev Arena it gains +89 Elo, reaching #4 (1769).
  4. Specs (per Artificial Analysis): 1M-token context window; text, image, video and speech input with text output.
  5. Staged rollout: Carbon is not publicly available yet. Starting this week Google is giving it to a subset of cyber defenders, then expanding access to trusted partners through its Fairwind Program, and it is taking part in the U.S. government's voluntary pre-release testing. Developer, business and consumer access is promised "as soon as possible".
  6. Pricing: the introductory price is $2 / $10 per 1M input/output tokens, with cached input 95% off. The standard price after that is $4 / $20.

Usage

How to get access

  • Now: access is limited to selected cybersecurity defenders and trusted partners through Google's Fairwind Program.
  • Later: Google says Carbon will come to developers (Gemini API), businesses and consumers through its product surfaces. Going by the Argon rollout, Ars Technica expects it to reach AI Ultra subscribers and Gemini API developers first.

Good fits

  • Vulnerability discovery, validation and automatic patching in security teams' workflows
  • Long-running engineering and agentic coding tasks
  • Full-stack web development and multimodal analysis

Notes

  • If you are choosing a model, compare cost per task with Gemini 4 Argon. Both have the same price, and Ars Technica describes Carbon as "essentially a slightly better version of Argon".
  • Google has not given an end date for the introductory pricing.