Google's new #1 model scores 59 on a benchmark, but only cyber defenders can use it
Google announced Gemini 4 Carbon on October 28 with a leaderboard-topping score and a price list. Access, for now, goes to a subset of security defenders.
Carbon posts 88.2% on SEC-bench Pro for finding vulnerabilities and 54.8% on ExploitGym for exploiting them, and Google has already priced it at $2 per million input tokens. The catch: it is rolling out through a trusted-partner programme first, with wider release promised "as soon as possible."
A reader who tried Google's own app found no sign of the previous model either.
A security-first launch with a price list but no public release
Google DeepMind's announcement calls Gemini 4 Carbon its "strongest model with state-of-the-art cybersecurity and long horizon capabilities." Access began this week for a subset of cyber defenders and will widen gradually to more trusted partners through the Fairwind Program, while Google keeps refining guardrails before a general release.
The numbers are the pitch. Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 with 59, scores 88.2% on SEC-bench Pro (vulnerability finding), 54.8% on ExploitGym (exploitation) and 71.0% on CWE-bench (patching). Google also reports 0.0% attack success on the Gray Swan indirect prompt-injection benchmark, against 0.7% for Gemini 4 Argon, and says the Chrome team measured more than 10x better patch generation than another lab's large frontier model.
Introductory pricing is $2 per million input tokens and $10 per million output tokens, rising to $4 and $20 later.
The thread: you can't use it, and you can't find the last one either
The loudest complaint was about availability. One reader described trying to follow Google's own trail:
"At this point if a Google PM handed me $100 and said "go figure out how to issue a prompt to Gemini 4 Argon" I'm only ~30% sure I could figure out how." — a Workspace user who could not find Argon in the Gemini app
Another opened the consumer site and saw older models only:
"When I open gemini.google.com I see 3.6 Flash and 3.1 Pro." — a reader checking the public Gemini site
A third, who spotted two bugs on the launch page itself (the benchmark link pointed to a page that did not list Carbon, and the screen reader read "[[duration]] minutes"), argued the announce-first pattern hands rivals a target:
"…now everyone knows what exactly they need to release eval results wise to beat Carbon." — a reader who caught the broken launch-page text
The same comment says Google fixed both bugs after it was posted.
What a score is worth before anyone can reproduce it
For a developer the practical read is narrower than the headline. Every figure above is Google's own or a leaderboard entry on a model that outsiders cannot yet call, so the first independent coding results will matter more than the launch post. Until then the only usable number is the price.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
Apple told developers to patch certificates by hand, then macOS 27.2 reverted the patch
Nine days after Apple was notified about the compromised eMudhra CAs, macOS 27.2 shipped and quietly undid the workaround users had applied.