Google is back with another power move in the AI race, dropping a brand-new frontier model when nobody expected it. Just two weeks after rolling out Argon, the tech giant officially announced Gemini 4 Carbon to push its software capabilities even further. It promises massive leaps in cybersecurity and coding, but there is 1 major catch keeping regular users from trying it today.
According to Google, Gemini 4 Carbon jumps straight to the number 1 spot on version 5.0 of the Artificial Analysis Intelligence Index. The new system outperforms rival tools like GPT-6 Astra and Claude Opus 5.5 across complex coding tasks, scientific reasoning, and full-stack web development.
Cybersecurity gains and prompt safety upgrades
The general reasoning scores for the new AI model look impressive. However, cybersecurity is where Gemini 4 Carbon makes its biggest technological leap. Internal testing shows the new model generates secure code at more than 2 times the success rate of Argon while patching system vulnerabilities 10X faster than competing labs. Security researchers at Grey Swan also reported that Carbon resisted 100% of indirect prompt-injection attacks, scoring a perfect 0.0% flaw rate during safety evaluations.
Because Carbon packs such advanced cyber defense tools, Google is taking a cautious approach to its public rollout. The company is working directly with the US government through a voluntary pre-release testing framework to vet guardrails before opening wider access. Select cybersecurity pros inside Google Fairwinds cyber defense program get initial access this week, while regular developers wait another 1 or 2 weeks.
API pricing details and public availability
When general access finally opens up, developers can tap into Carbon through standard Gemini API connections. Google set an introductory rate of $2 per million input tokens and $10 per million output tokens. This matches the initial promotional pricing used for Argon. Cached input tokens will also receive a massive 95% discount to help teams run complex multi-step engineering workflows without blowing their budgets.
Releasing 2 frontier AI models in less than 30 days shows how fast Google wants to regain momentum in the tech sector. While waiting for public access can feel frustrating, testing advanced cyber capabilities behind closed doors remains a smart safety play.
The post Google Announces Gemini 4 Carbon with Massive Upgrades, But You Can’t Use It Yet appeared first on Android Headlines.
​Â
