From the source
Google DeepMind is announcing Gemini 4 Carbon, a frontier model it says is its strongest yet for cybersecurity and long-horizon tasks, following Gemini 4 Argon two weeks earlier.
- Carbon ranks first on v5.0 of the Artificial Analysis Intelligence Index at 59 and leads SEC-bench Pro at 88.2% for finding vulnerabilities, ExploitGym at 54.8% for exploitation and CWE-bench at 71.0% for patching.
- The company says internal benchmarks show more than 2x uplift over Argon in generating secure code, and that the Chrome team reported Carbon produced high-quality patches more than 10X better than another lab's large frontier model.
- Rollout is deliberately staged: from this week Carbon goes to a subset of cyber defenders, with access widened gradually to trusted partners through the Fairwind Program before it reaches developers, businesses and consumers.
- Carbon launches at the same introductory price as Argon, $2 per million input tokens and $10 per million output tokens, with cached input tokens at 95% off the input price.
- Google says Carbon builds on Argon's safety stack and cut successful cheating on Terminal-Bench 4.0 to 0.0%, with high-severity misalignment alerts in internal live coding environments down almost 80%.
