Google has introduced Gemini 4 Carbon, a new AI model it calls its strongest for cybersecurity and long-running engineering work. Access is limited for now to vetted cyber defenders through Google’s Fairwind program, while the company completes voluntary U.S. government testing.
Gemini 4 Carbon: key facts
- Built for multi-day coding tasks and finding and fixing software vulnerabilities.
- Scores 88.2% on SEC-bench Pro, which tests whether AI can find real vulnerabilities.
- Scores 71% on CWE-bench, which measures whether it can produce working fixes.
- Scores 59 on the Artificial Analysis Intelligence Index v5, which one report places first, ahead of Claude Opus 5.5 and GPT-6 Astra.
- Introductory pricing is reported at $2 per million input tokens and $10 per million output tokens.
Who can use it and when
Only selected security teams can use Carbon today, through the Fairwind program. Google has not announced a date for broad availability. It says it is taking part in voluntary U.S. government pre-release testing before any wider rollout.
The restricted approach reflects the dual-use nature of the technology: a model that is good at finding software flaws can help defenders patch them, but it could also help attackers. Limiting early access to vetted defenders is Google’s way of managing that risk.
Carbon follows Gemini 4 Argon, an earlier model in the same family that also went first to security teams. Some early commentary describes Carbon as an incremental step, with a lead of only a few points over rival models on several benchmarks. Note that most of these scores are published by Google or by benchmark providers and have not been independently verified by this site.
Alphabet earnings and the Gemini app
The launch came the same week as Alphabet’s third-quarter results. Reports say revenue was $127.58 billion, slightly below analyst expectations, and shares fell more than 7%. Google Cloud revenue more than doubled to $30.7 billion, and the Gemini app reached 1 billion users, according to the reports. Alphabet also said annual capital spending would top $200 billion.
Why it matters
The AI race is increasingly about chips, cybersecurity and long autonomous tasks, not only chatbots. For background on the hardware side, read our report on OpenAI’s reported plan to buy Google TPU chips, and see our look at the new Apple M6 iMac and MacBook Pro for another big tech launch this week.
Sources: The Neuron, KillTheAI, Decrypt
Leave a Reply