Google on Wednesday unveiled Gemini 4 Carbon, which it calls its strongest model, with a focus on cybersecurity and long-horizon engineering tasks, according to Google’s announcement and Reuters. Reuters described the launch as stepping up competition with OpenAI and Anthropic. Google’s announcement says Carbon builds on Gemini 4 Argon, which it describes as launched “two weeks ago.” Google’s Argon post is dated Oct. 13 and made that model available to developers and enterprises. AI Tech Daily reported Sept. 30 that Google had first unveiled Argon in a phased rollout to trusted cyber partners.
Google said Carbon ranks first on version 5.0 of the Artificial Analysis Intelligence Index with a score of 59. Artificial Analysis confirmed the score and the No. 1 spot on its own site, and its index page lists Claude Opus 5.5 (max) at 58. Artificial Analysis said Carbon has the top score on four of the index’s nine evaluations and is not leading in coding and knowledge work. It also placed Carbon second to Claude Opus 5.5 on its Coding Agent Index, 59 to 62. Artificial Analysis said Carbon, like Argon at launch, is being rolled out to selected users and is not yet publicly available.
On external cybersecurity benchmarks, evaluated without guardrails, Google reported 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on Collinear’s CWE-bench for patching. Google also said Carbon more than doubles Argon’s success rate at generating secure code on internal benchmarks, reduced successful cheating on Terminal-Bench 4.0 to 0.0%, and posted 0.0% attack success on Gray Swan’s indirect prompt injection benchmark. All of those figures are Google’s own reporting.
Access is limited. Google said it will make Carbon available this week to a subset of cyber defenders and will gradually phase in additional trusted partners through its Fairwind Program. Google said it is “actively engaged” in the U.S. government’s voluntary process for pre-release model testing and is gathering feedback and iterating on guardrails before making Carbon available to developers, businesses and consumers “as soon as possible.” It gave no date. Reuters said Carbon is being made available first to cybersecurity defenders as the company continues testing safeguards.
Google said Carbon will launch at the same introductory price as Argon: $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off the input price. A footnote in Google’s post says $4 and $20 per million tokens will apply after the introductory period expires, and the post gives no end date for that period. Artificial Analysis said Google is offering a 50% launch discount with no confirmed end date, and that its charts use the standard $4 and $20 pricing.
Sources
- Google (Koray Kavukcuoglu): Gemini 4 Carbon: our next frontier model (October 28, 2026)
- Google: Gemini 4 Argon: Built to solve your hardest problems (October 13, 2026)
- Reuters: Google launches Gemini 4 Carbon AI model with cybersecurity focus (October 28, 2026; U.S. News mirror)
- Artificial Analysis: Gemini 4 Carbon is the new #1 on the Artificial Analysis Intelligence Index v5 (October 28, 2026)
- Artificial Analysis Intelligence Index v5.0 page