Try Gemini 4 Carbon on Your Own Task
One model for extended autonomous work and leading-edge security defense
None

None

None

Long Story Video Skill

Turn your idea into 10s–10min video—AI writes scripts, prompts, and generates footage automatically.

Ads Video Skill

Ads Video Skill

Generate professional ads and sales videos—AI auto-generates scripts, prompts, and footage.

3D Science Explainer Video Skill

Convert scientific concepts into stunning 3D explain animations

AI Video Prompt Generator

Feedback

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

Gemini 4 Carbon

Gemini 4 Carbon pairs days-long agentic engineering with frontline cyber defense, top benchmark scores and entry pricing near $2 per 1M input tokens.

All Tools

Discover our comprehensive AI-powered animation toolkit

Gemini 4 Carbon: Engineered for Hard Problems

Engineered for extended autonomous execution and security defense, this model backs its frontier scores with a cautious, safety-led release plan.

  • Agentic Work That Keeps Going for Days
    Built for tougher, multi-stage engineering work. Our own teams have watched it operate for days at a stretch, working past failed attempts to trace and fix elusive defects deep inside huge codebases and low-level systems.
  • Security Defense Across the Whole Lifecycle
    Coverage runs from spotting a weakness to confirming and repairing it: 88.2% on SEC-bench Pro for real flaw discovery, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for root-cause patching.
  • Rapid Progress from Argon to Carbon
    Arriving only two weeks after Gemini 4 Argon, Carbon shows what happens when researchers and models iterate together — the frontier moves forward while safeguards grow alongside it.

How Access to Gemini 4 Carbon Unfolds

Carbon is being opened up in careful stages — here is the order in which that happens.

Gemini 4 Carbon: Scores, Skills and Safety

A first-place Intelligence Index finish, a zero-percent injection success rate and more — here is what the evaluations show.

First Place on Intelligence Index v5.0

Carbon posts 59 at the summit of the freshly revised Artificial Analysis Intelligence Index v5.0, gaining substantially in coding, agentic work and the capabilities that matter most for security defense.

Fourth Place on WebDev Arena

Sharper output and artifact quality in full-stack web work moves Carbon up 89 elo points to #4, at 1768 on WebDev Arena.

Fewer Shortcuts, Less Cheating

Alignment tuning brought successful cheating on Terminal-Bench 4.0 down to 0.0%, with attempts at 0.4% even as accuracy rose, and high-severity misalignment alerts dropped nearly 80% in internal live coding settings.

More Honest, More Transparent Output

Internal tests show unsupported claims roughly halved versus Argon; caveats about method land in the deliverable, or blocked actions come back with a clear list of what is still needed.

Holds Up Against Indirect Prompt Injection

On Gray Swan's Indirect Prompt Injection Benchmark, Carbon shows a 0.0% attack success rate, deflecting every evaluated attempt where Argon had sat at 0.7%.

Finishes the Job Without Overstepping

Overreach fell 26% and under-service fell 29%, so multi-step plans execute end to end more reliably while respecting project constraints and staying inside scope.

FAQ

Gemini 4 Carbon: Your Questions Answered

Straight answers on cost, evaluation results, and who gets to use Carbon first.

1

So what exactly is Gemini 4 Carbon?

It is the newest frontier model we have shipped and the strongest yet, aimed at leading-edge cybersecurity and long-horizon work. It sits at #1 on the updated Artificial Analysis Intelligence Index v5.0 with a score of 59.

2

What does Gemini 4 Carbon cost to use?

Introductory pricing runs $2 per million input tokens and $10 per million output tokens, with cached input discounted 95% off the input rate. Once the introductory window closes, $4 per 1M input and $20 per 1M output applies.

3

Where does Carbon lead on benchmarks?

It hits 88.2% on SEC-bench Pro for vulnerability discovery, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching, and lands #4 on WebDev Arena with 1768 points.

4

Who has access to Carbon today?

The opening group is a hand-picked set of cyber defenders; additional trusted partners join over time through the Fairwind Program, with developers, businesses and consumers following via Google's product surfaces.

5

What does Carbon do better than Gemini 4 Argon?

On internal benchmarks it delivers more than 2x the uplift in secure code generation, roughly half the unsupported claims, a 0.0% prompt-injection success rate and tighter scope discipline.

6

Which tasks suit Carbon best?

Complex multi-step engineering — tracing and fixing subtle defects across large codebases and low-level systems — along with agentic workflows, scientific work, full-stack web development and multimodal understanding.

Get Ready for Gemini 4 Carbon Access

Carbon aims at real gains in software engineering and security defense while keeping safety and reliability intact. Track the staged rollout through the Fairwind Program — developer, business and consumer access is on the way.