Gemini 4 Carbon
Google's strongest Gemini model, in a staged release to cyber defenders
Context window
Not announced
Input / 1M tokens
$2.00
Output / 1M tokens
$10.00
Provider
Google's announced introductory price, the same as Gemini 4 Argon's: $2/$10 per 1M input/output tokens, with cached input 95% off the input price. After the introductory period: $4/$20 per 1M; Google has not given an end date. Carbon is not yet listed on the Gemini API pricing page. Artificial Analysis lists a 1M-token context window, which Google has not confirmed. · Data verified 2026-11-03
Gemini 4 Carbon (October 28, 2026) is Google's next frontier model after Gemini 4 Argon, which Google calls its strongest model for cybersecurity and long-horizon work. Google reports it ranks first on the Artificial Analysis Intelligence Index v5.0 with 59 and sets new highs on SEC-bench Pro and ExploitGym, with the cybersecurity results measured without guardrails. Access started with a subset of cyber defenders and is expanding gradually to trusted partners through Google's Fairwind Program, and the model is in the US government's voluntary pre-release testing. Google says developers, businesses and consumers will get it soon but has not given a date.
How to access it
Not generally available. Since October 28, 2026 Google has been rolling Carbon out to a subset of cyber defenders, expanding gradually to trusted partners through its Fairwind Program. Google says developer, business and consumer access will follow but has not given a date or published an API model ID.
Strengths
- ✓Ranked #1 on the Artificial Analysis Intelligence Index v5.0 with a score of 59 (per Google)
- ✓88.2% on SEC-bench Pro vulnerability finding and 54.8% on ExploitGym (Google-reported)
- ✓0.0% attack success rate on Gray Swan's indirect prompt injection benchmark (Google-reported)
- ✓Google reports about half as many unsupported claims as Gemini 4 Argon
- ✓Announced at the same $2/$10 introductory price as Argon
Best for developers who...
When to choose it (and when not to)
Reach for Gemini 4 Carbon when...
- →You are an approved cyber defender or trusted partner in Google's Fairwind Program
- →You want Google's strongest model for vulnerability research and patching
Look elsewhere if...
- ✕You need API or app access today (no general release date yet)
- ✕You need a confirmed context window or model ID (Google has not published either)
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| Artificial Analysis Intelligence Index v5.0 | 59 | Ranked #1, per Google; Artificial Analysis lists the same score |
| SEC-bench Pro | 88.2% | Vulnerability finding, Google-reported, evaluated without guardrails |
| ExploitGym | 54.8% | Exploitation, Google-reported, evaluated without guardrails |
| CWE-bench | 71.0% | Patching, Google-reported |
| WebDev Arena | 1768 Elo | Ranked #4, per Google |
| Gray Swan Indirect Prompt Injection | 0.0% | Attack success rate (lower is better), Google-reported; Gemini 4 Argon 0.7% |
Compare Gemini 4 Carbon
Compare Gemini 4 Carbon with any other model
Build a comparison →All model comparisons →