{
  "id": 17015431,
  "title": "Google's Gemini 4 Carbon looks like a diamond in the rough",
  "url": "https://urgent.news/2026/10/29/googles-gemini-4-carbon-looks-like-a-diamond-in-the-rough",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-29T05:29:00.000Z",
  "source": {
    "name": "The Register Science",
    "slug": "the-register-science",
    "url": "https://www.theregister.com/ai-and-ml/2026/10/29/googles-gemini-4-carbon-looks-like-a-diamond-in-the-rough/5306149"
  },
  "original_language": "en",
  "account": "Just two weeks after unveiling Gemini 4 Argon, Google has unveiled an enhanced version of the model, named Gemini 4 Carbon, slated for release at the same cost initially. Koray Kavukcuoglu, chief AI architect at Google, affirmed Gemini 4 Carbon as the company's most formidable model, especially in cybersecurity and extended tasks. The name 'Carbon' bears an ironic twist, considering Google's commitment to a carbon-neutral future. If Google's branding were more consistent across the board, 'Gemini 4 Americium' might better encapsulate the company's ethos. Another potential naming opportunity was 'Gemini 4 Silicon,' but the company missed the chance to use Si, a reference to 'Super Intelligence.' Initially priced at $2 per million input tokens and $10 per million output tokens, Gemini 4 Carbon will cost the same as its predecessor, Argon, during a temporary introductory phase. Post the introductory period, Google plans to increase the rates to $4 per million input tokens and $20 per million output tokens, mirroring Anthropic's Claude Opus 5.5, which the tech giant considers a direct competitor. Google's DeepMind AI division faced skepticism about its leadership after the recent change, but Gemini 4 Carbon has delivered impressive benchmark scores, outperforming competitors in SEC-bench Pro, ExploitGym, and CWE-bench, while securing the fourth position on WebDev Arena, trailing behind Opus 5.5, GPT-6.1 Astra, and Sonnet 5.5. Currently, it ranks second, behind Opus 5.5, on Arena.ai's leaderboard for Best Overall Agent. A model rating service, Artificial Analysis, places Google's new model at the top of its Artificial Analysis Intelligence Index v5, narrowly missing out to Claude Code Opus 5.5 on the Coding Agent Index. Koray Kavukcuoglu highlighted Gemini 4 Carbon's robust cybersecurity features, which will be accessible to a select group of cybersecurity professionals this week and subsequently through the company's Fairwind access program. On internal benchmarks, Carbon demonstrates a remarkable 2x improvement in the success rate for generating secure code over Argon, and it outperforms other large frontier models by over 10x when generating high-quality patches. Google's CEO, Sundar Pichai, announced, \"Gemini 4 Argon and Carbon are now leading at the frontier,\" in a statement accompanying Alphabet's Q3 2026 earnings, emphasizing that their first-party model APIs are now processing nearly 32 billion tokens per minute, a fourfold increase from a year ago. Despite Google's aspiration to surpass Anthropic and OpenAI in terms of model performance and usage, the company already leads in revenue. Nate Elliott, an Emarketer principal analyst, commented on Google's strong Q3 earnings, saying, \"In a sly nod to the way AI-only companies talk about their revenue, Google bragged about its half-trillion dollar run rate – almost ten times more than either OpenAI or Anthropic.\" Google plans to make Gemini 4 Carbon available to a wider audience as soon as possible.",
  "summary": "Undercutting Claude on price ahead of Anthropic's IPO",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}