Google has completely surpassed ChatGPT and Claude by unveiling Gemini 4 Carbon — a model that beats all leading…

Google has completely surpassed ChatGPT and Claude by unveiling Gemini 4 Carbon — a model that beats all leading competitors on most tests. — The system took first place in the Artificial Analysis V5 ranking and improved significantly in programming, cybersecurity, and handling complex multi-step tasks. — The AI reached 71% on CWE-bench, 65% on Terminal-Bench 4.0, and 96% on GraphWalks when working with context up to 1 million tokens. — Carbon also earned $16,484 on Vending-Bench 2, outperforming GPT-6 Astra and Claude Opus 5.5. At the same time, the developers kept the previous pricing — $2 per million input tokens and $10 per million output tokens. I really want to try this model already.