Gemini 4 Carbon
Google announced Gemini 4 Carbon, its strongest model, with top cybersecurity and long-horizon results, available first to a select group of cyber defenders before expanding to other Fairwind Program partners.
Carbon follows Gemini 4 Argon, launched two weeks earlier, and Google says it more than doubles Argon's success rate at generating secure code on internal benchmarks. It is built for long agentic tasks, and Google says internal teams saw it run for days on multi-step engineering problems. Access starts with a small group of cyber defenders and widens to more trusted partners, while Google iterates on guardrails before opening it to developers, businesses and consumers.
- Date
- Wednesday 28 October 2026
- Lab
- Google DeepMind
- Kind
- model
- Access
- research preview
- Price
- $2 per million input tokens and $10 per million output tokens (introductory)
Figures
| Measure | Value | Measured by |
|---|---|---|
| SEC-bench Pro | 88.2% Google calls it state-of-the-art; evaluated without guardrails | company |
| Secure code generation success rate vs Argon | more than 2x internal benchmarks | company |
| Input token price | $2 per million tokens introductory price, same as Argon | company |
| Output token price | $10 per million tokens introductory price, same as Argon | company |
Access is limited to select cyber defenders for now. General availability to developers and consumers has no stated date.
Sources
- blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbo
- deepmind.google/blog/gemini-4-carbon-our-next-frontier-model/
- theverge.com/tech/1021339/google-gemini-4-carbon-cybersecurity-frontier
This record was checked and corrected against its sources on 28 October 2026. How we check