Home · Technology · Oct 30 archive
Google Launches Gemini 4 Carbon for Enhanced Cybersecurity
· Confirmed
In Short: Google has unveiled Gemini 4 Carbon, its latest AI model designed to excel in cybersecurity and complex coding tasks.
Google announced the launch of Gemini 4 Carbon, its newest AI model, on October 28, 2026. The model is designed to enhance cybersecurity and handle complex coding projects, building on the capabilities of Gemini 4 Argon, which was introduced two weeks earlier.
According to Google, Gemini 4 Carbon leads in several cybersecurity benchmarks, including SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching. The model also excels in coding tasks, making it the best available for cybersecurity work, according to Google.
Google CEO Sundar Pichai announced the model, highlighting its perfect resistance to indirect prompt injection and its ability to significantly reduce unsupported claims. In internal testing, Carbon cut unsupported claims roughly in half compared to Argon, demonstrating improved alignment and security.
Gemini 4 Carbon will initially be offered at the same introductory price as Argon: $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off the input token price. After the introductory period, the cost will increase to $4 per million input tokens and $20 per million output tokens.
The model's performance in cybersecurity tasks is notable, with Google reporting that Carbon achieves over double the success rate of Argon in generating secure code. Additionally, Google's Chrome team found Carbon to be more than 10 times better than another lab's large model at producing high-quality patches, and cloud security company Wiz reported it leading on web penetration testing.
On WebDev Arena, Carbon gained 89 Elo points to reach fourth place with a score of 1,768, reflecting improvements in the quality of web development outputs and artifacts. This indicates that Carbon is not only strong in cybersecurity but also excels in web development tasks.
Google's announcement of Gemini 4 Carbon comes amid growing competition among major AI developers to build models capable of performing increasingly sophisticated software engineering and cybersecurity work. The company says Carbon reflects the pace at which it is pushing the frontier forward, coming with a significant leap in cyberdefense.
In live coding environments with internal teams, Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with cheating attempts falling to 0.4% while overall accuracy improved. High-severity misalignment alerts dropped by almost 80%, demonstrating the model's improved safeguards.
What this adds
On Gray Swan’s Indirect Prompt Injection Benchmark, Carbon recorded a reported attack success rate of 0%, compared with Argon’s 0.7%. This further underscores Carbon's enhanced security features.
The company reported that Carbon improves collaboration and can follow through on tasks “without overstepping,” indicating a more reliable and aligned model compared to its predecessor.
Background
Gemini is already available globally on Android Auto and vehicles with Google built-in, but Personal Intelligence takes things further by connecting information from Google services that drivers already use.
Google Chrome is rolling out a new autofill feature called 'Suggested by Gemini' for users with Personal Intelligence enabled.
What's confirmed
- It shows significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense, landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and CWE-bench (71.0%) for patching.
- Carbon ranks #1 on the recently updated v5.0 of Artificial Analysis Intelligence Index at 59.
- Gemini 4 Carbon will launch at the same introductory price 1 as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.
- Carbon achieves over double the success rate of Gemini 4 Argon in generating secure code according to internal benchmarks.
- Google has announced Gemini 4 Carbon, its latest frontier AI model, building on Gemini 4 Argon, which launched two weeks ago.
- Google’s Chrome team found it more than 10 times better than another lab’s large model at producing high quality patches, and cloud security company Wiz reported it leading on web penetration testing.
What's still developing
- In particular, we made significant strides in model alignment with Argon, and reported its propensities to behave in an aligned way in our Safety and Security report.
- The company introduced Gemini 4 Carbon on Wednesday, calling it its strongest model yet, and said a small group of cyber defenders will get access first while it tests the safeguards.
Sources
- blog.googlelink
- Basic Tutorialslink
- Brave New Coinlink
- Dataconomylink
- Latest news from Azerbaijanlink
- Online Tech Tipslink
- Technology Orglink
- The Vergelink
- Ticker Newslink
- Fonearenalink
- Rollingoutlink
- Nerdschalklink
- Traklink
- Unitelink
- WarpBeat — background on Google Rolls Out Personalized Gemini for In-Car Use link
- WarpBeat — background on Google Chrome Launches 'Suggested by Gemini' Autofill link
- Telusko — video link