Home · Technology · Oct 31 archive
Google Launches Gemini 4 Carbon with Enhanced Cybersecurity and Coding Capabilities
· Confirmed
In Short: Blog.google and 9To5Google highlighted that Carbon ranks #1 on the recently updated v5.0 of the Artificial Analysis Intelligence Index at 59, showing significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense.
According to blog.google and 9To5Google, Google today announced the launch of Gemini 4 Carbon, the latest iteration of its AI model designed to excel in cybersecurity and complex coding tasks. Both publications reported that Carbon will be available at the same introductory price as Argon, with $2 per million input tokens and $10 per million output tokens, and cached input tokens priced at 95% off the input token price.
Brave New Coin and FoneArena.com reported that Google CEO Sundar Pichai announced Gemini 4 Carbon as the newest frontier model in the Gemini 4 family. Both outlets noted that Carbon targets demanding technical and security workloads, while Genie 4 aims to generate dynamic digital worlds that respond to user instructions in real time.
Blog.google and 9To5Google highlighted that Carbon ranks #1 on the recently updated v5.0 of the Artificial Analysis Intelligence Index at 59, showing significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense. Both publications noted that Carbon achieved an 88.2% success rate in SEC-bench Pro for finding vulnerabilities, a 54.8% success rate in ExploitGym for exploitation, and a 71.0% success rate in CWE-bench for patching.
According to blog.google and 9To5Google, Carbon will be made available to a subset of cyber defenders starting this week, with gradual phase access to additional trusted partners through Google's Fairwind Program. Both outlets noted that Google is continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers.
Blog.google and 9To5Google reported that Carbon drives a strong uplift in generating secure code and shows reduced misalignment and fabrication rates in live coding environments with Google's own teams. Both publications noted that Carbon achieved an 89-point improvement on WebDev Arena, where it ranked fourth with a score of 1,769.
Carbon reduced unsupported claims by approximately half compared with Argon, while high-severity misalignment alerts in internal coding environments fell by nearly 80 per cent. Both outlets noted that Carbon is tailored for complex engineering and coding challenges.
Blog.google and 9To5Google noted that Carbon achieved over double the success rate of Gemini 4 Argon in generating secure code according to internal benchmarks. Both publications highlighted that Carbon is designed for harder, long-horizon engineering problems and built for sustained agentic workflows.
Blog.google and 9To5Google reported that Carbon reduced successful cheating on Terminal-Bench 4.0 to 0.0%, with cheating attempts falling to 0.4% while overall accuracy improved. Both outlets noted that high-severity misalignment alerts dropped by almost 80% in live coding environments with internal teams.
Blog.google and 9To5Google agreed that Carbon will be available at the same introductory price as Argon, with $2 per million input tokens and $10 per million output tokens, and cached input tokens priced at 95% off the input token price. Both publications noted that Carbon will be made available to a subset of cyber defenders starting this week, with gradual phase access to additional trusted partners through Google's Fairwind Program.
Both publications agreed that Carbon showed significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense, achieving an 88.2% success rate in SEC-bench Pro for finding vulnerabilities, a 54.8% success rate in ExploitGym for exploitation, and a 71.0% success rate in CWE-bench for patching.
Both publications agreed that Carbon reduced unsupported claims by approximately half compared with Argon, while high-severity misalignment alerts in internal coding environments fell by nearly 80 per cent.
Both publications agreed that Carbon achieved an 89-point improvement on WebDev Arena, where it ranked fourth with a score of 1,769.
What this adds
The new model, Gemini 4 Carbon, is designed to excel in cybersecurity and complex coding tasks, but anyone planning a migration or procurement cycle around Carbon today is planning around a promise, not a ship date, according to reports.
Background
Google announced Gemini 4 Carbon, its latest AI model designed to excel in cybersecurity and complex coding tasks.
Google today launched Gemini 4 Carbon, its latest artificial intelligence model, focusing on cybersecurity and complex coding tasks.
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.
- Our researchers and models are working hand in hand to advance capabilities faster, while strengthening our safeguards in tandem.
- Building on Gemini 4 Argon, which was launched two weeks ago, Carbon illustrates the pace at which we’re pushing the frontier forward.
- Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.
- 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.
- Google announced Gemini 4 Carbon on October 28, describing it as its strongest artificial intelligence model, with improved cybersecurity capabilities and the ability to handle complex, long-running tasks.
- Google CEO Sundar Pichai announced Gemini 4 Carbon as the newest frontier model in the Gemini 4 family.
- Google has announced Gemini 4 Carbon, its latest frontier AI model, building on Gemini 4 Argon, which launched two weeks ago.
- Carbon achieves over double the success rate of Gemini 4 Argon in generating secure code according to internal benchmarks.
What's still developing
- 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.
- Anyone planning a migration or procurement cycle around Carbon today is planning around a promise, not a ship date.
- The same capabilities that help a team find and fix a vulnerability can help an attacker exploit it.
- Those restrictions limit malicious use while Anthropic works to cut false positives in secure coding.
- Anthropic says the model can recognize when a user is running a cybersecurity test, making it hard to measure how safe it really is in real-world use.
- The release purportedly offers Mythos-like cybersecurity and biology capabilities.