Google Introduced Powerful Gemini 4 Carbon Model
Google introduced the Gemini 4 Carbon model, which demonstrates outstanding results compared to its counterparts. The new solution outperforms competitors in most benchmarks.
Artificial Intelligence took first place in the Artificial Analysis V5 ranking. The model showed strong skills in programming, cybersecurity, and solving complex multi-stage tasks.
In security tests on CWE-bench, the result was 71%, and on Terminal-Bench 4.0 — 65%. When working with a context of up to 1 million tokens, accuracy reached 96%.
The model earned $16,484 on Vending-Bench 2. This indicator is higher than the results of GPT-6 Astra and Claude Opus 5.5.
You can test the model's capabilities through Google's Antigravity service.









