Contra - A professional network for the jobs and skills of the futureGemini 4 Carbon Rollout and Cybersecurity Benchmark Results
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
88.2%. That is Google's newest model's score at finding real vulnerabilities on SEC-bench Pro. And you cannot have it yet.
Gemini 4 Carbon landed today, and the rollout order is the story.
→ This week: a subset of cyber defenders. → Next: trusted partners, in phases. → Later: developers, businesses, consumers, "as soon as possible."
The chart tells you why. Find 88.2%, exploit 54.8%, patch 71.0%. A model that finds holes better than it fixes them goes to the people fixing things first.
Two more lines I read twice: → Prompt-injection attack success on Gray Swan's benchmark: 0.0%, down from 0.7% on Argon two weeks ago. → Price: $2 in, $10 out per million tokens during the intro period, then $4 and $20.
I run a one-person company on AI agents, so here is what changes for me:
Rule 1: Every vendor pitch of "finds vulnerabilities" now has a public bar to clear: 88.2. Rule 2: A 0.0% injection score is a test result, not a guarantee. My approval gate stays. Rule 3: Budget at the post-intro price. The doubling is in the footnote.
The strongest model of the year is, for now, a defender's tool. That ordering is new.
Agree?
Post image
Back to feed
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started