Google chorshanba kuni Gemini 4 Carbon — kompaniya ma'lumotiga ko'ra, zaifliklarni aniqlash, ulardan foydalanish va tuzatish bo'yicha benchmarklarda raqib modellarni ortda qoldiradigan chegara sun'iy intellekt modelini e'lon qildi.
Model hali keng tarqatilmagan. Bu hafta Google kiber himoyachilarning tanlangan guruhiga kirish huquqini ochadi, so'ngra uni bosqichma-bosqich Fairwind dasturi orqali kengroq tasdiqlangan hamkorlar doirasiga kengaytiradi.
Fairwind hukumatlarga, muhim infratuzilma operatorlariga va yirik texnologiya platformalariga kiber himoya uchun ilg'or modellarga erta kirish huquqini taqdim etadi. Kompaniya ma'lumotiga ko'ra, hozirda 650 dan ortiq tashkilot ishtirok etmoqda.
Carbon ikki hafta oldin Google ishga tushirgan Gemini 4 Argondan keyin keladi. Ba'zi Fairwind hamkorlari allaqachon Argonning kiberxavfsizlik himoya imkoniyatlaridan eksklyuziv foydalanish huquqiga ega.
Google ma'lumotiga ko'ra, Carbon SEC-bench Pro'da 88,2% natija ko'rsatdi — bu modelning haqiqiy dunyo dasturiy ta'minotidagi kamchiliklarni aniqlash qobiliyatini sinovdan o'tkazadi. Shuningdek, u ExploitGym ekspluatatsiya benchmarkida 54,8% va Collinear'ning CWE-bench tuzatish benchmarkida 71,0% natija qayd etdi, kompaniya taqqoslashida OpenAI'ning GPT-6 Astra va Anthropic'ning Claude Opus 5.5 hamda Claude Fable 5.1 modellaridan yuqori natija ko'rsatdi.
Kompaniya ta'kidlaganidek, bu natijalar hech qanday cheklovlarsiz olingan.
Google ekspluatatsiyani himoya qadam sifatida ko'radi, bu jamoalarga nimani tuzatishni hal qilishga yordam beradi. "Qaysi zaifliklarni birinchi navbatda tuzatish yoki umuman tuzatish kerakligini aniqlash uchun kiber himoyachilar tobora ko'proq modellarga qaysi zaifliklardan foydalanish mumkinligini tekshirish uchun tayanmoqda", — dedi kompaniya.
Google'ning ichki benchmarklariga ko'ra, Carbon xavfsiz kod yozishda Argondan ikki baravar muvaffaqiyatliroq. Yuqori sifatli tuzatmalarni yaratish borasida Google'ning Chrome jamoasi Carbon raqib laboratoriyadan yirik chegara modelidan 10 baravardan ortiq ustun ekanligini aniqladi.
Google'ga tegishli bulut xavfsizlik firmasi Wiz Carbon'ning veb penetratsiya sinovlari benchmarklarida birinchi o'rinni egallaganini kuzatdi.
Xavfsizlik frontida Gray Swan'ning bilvosita prompt in'ektsiya benchmarkida Carbon'ga qarshi hech qanday hujum muvaffaqiyatli bo'lmadi, Argon uchun esa bu ko'rsatkich 0,7% edi. Terminal-Bench 4.0 da Carbon'ning hech bir aldash urinishlari muvaffaqiyatli bo'lmadi va bu urinishlar 0,4% gacha tushdi.
Argon bilan solishtirganda, yangi model taxminan yarmiga kamroq asossiz da'volar qildi. Google'ning ichki jamoalari Carbon'ni haqiqiy kodlash muhitlarida ishlatib, yuqori darajadagi mos kelmaslik ogohlantirishlarida deyarli 80% pasayishni kuzatdi.
Google AQSh hukumatining modellarni chiqarishdan oldin sinovdan o'tkazish bo'yicha ixtiyoriy sa'y-harakatlarida ishtirok etayotganini aytdi. "Biz Carbon'ni ishlab chiquvchilar, bizneslar va iste'molchilarga mahsulot yuzalarimiz orqali iloji boricha tezroq taqdim etishdan oldin fikr-mulohazalarni yig'ish va cheklovlarni takomillashtirishni davom ettirmoqdamiz", — dedi kompaniya.
Carbon Argonning kirish narxlarida — million kirish tokeni uchun 2 dollar va million chiqish tokeni uchun 10 dollar — ishga tushadi. Kirish davri tugagach, narx mos ravishda 4 va 20 dollarga ko'tariladi.
Bog'liq: Cisco AI agent hujumlarini zararsiz qadamlar ichida yashirishni aniqlash uchun Quarry'ni ochiq manbaga aylantirdi
Bog'liq: O'g'irlangan AI mulohazalari Xitoyga arzonroq raqiblar yaratishda yordam berishi mumkin
Bog'liq: Mythos 5 iyul hujumida haqiqiy dunyo zararini e'tiborsiz qoldirdi
Google on Wednesday announced Gemini 4 Carbon, a frontier AI model that, according to the company, beats rival models on benchmarks for detecting, exploiting and patching vulnerabilities.
The model is not broadly available yet. This week, Google is opening access to a select group of cyber defenders, then expanding it in stages to a wider pool of approved partners via its Fairwind Program.
Fairwind offers governments, critical infrastructure operators and major technology platforms early access to advanced models for cyber defense. The company says more than 650 organizations currently participate.
Carbon follows Gemini 4 Argon, which Google launched two weeks ago. Some Fairwind partners already have exclusive access to Argon’s cybersecurity defense capabilities.
Google said Carbon scored 88.2% on SEC-bench Pro, which tests a model’s ability to uncover flaws in real-world software. It also achieved 54.8% on the ExploitGym exploitation benchmark and 71.0% on Collinear’s CWE-bench patching benchmark, topping OpenAI’s GPT-6 Astra and Anthropic’s Claude Opus 5.5 and Claude Fable 5.1 in the company’s comparison.
These results were obtained with no guardrails in place, the company noted.
Google frames exploitation as a defensive step that helps teams decide what to fix. “To prioritize which vulnerabilities to remediate first or remediate at all, cyber defenders increasingly rely on models to check which vulnerabilities are exploitable,” the company said.
Based on Google’s internal benchmarks, Carbon is more than twice as successful as Argon at writing secure code. When it came to generating high-quality patches, Google’s Chrome team found Carbon outperformed a large frontier model from a rival lab by more than 10 times.
Wiz, the Google-owned cloud security firm, has observed Carbon taking the top spot on web penetration testing benchmarks.
On the safety front, no attack succeeded against Carbon in Gray Swan’s indirect prompt injection benchmark, compared with a 0.7% success rate for Argon. On Terminal-Bench 4.0, none of Carbon’s cheating attempts succeeded, and those attempts fell to 0.4%.
Compared with Argon, the new model made about half as many unsupported claims. Google’s internal teams, using Carbon in real coding environments, saw a nearly 80% decrease in high-severity misalignment alerts.
Google says it is involved in the US government’s voluntary effort to test models before they are released. “We’re continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible,” the company said.
Carbon will launch at Argon’s introductory rates of $2 per million input tokens and $10 per million output tokens. When the introductory period ends, the price goes up to $4 and $20, respectively.
Related: Cisco Open Sources Quarry to Catch AI Agent Attacks Hidden in Harmless Steps
Related: Stolen AI Reasoning Could Help China Build Cheaper Rivals
Related: Mythos 5 Ignored Real-World Harm in July Attack
