AI News Daily한국어
Sunday, November 1, 2026

Story 5 of 5

Gemini 4 Carbon tops the intelligence index but only cyber defenders can use it; users ask how soon "soon" is

First seen on X 16 hours ago@_mohansolo ♥ 387

Gemini 4 Carbon tops the intelligence index but only cyber defenders can use it; users ask how soon "soon" is

Gemini 4 Carbon, which Google announced on Oct. 28 (US time), is still open only to some cyber defense organizations, so the question circulating is when it will be released rather than how it performs.

Per 9to5Google, Carbon follows Gemini 4 Argon by two weeks. Google cited "significant gains" in coding, agentic tasks and defense. It ranks #1 on Artificial Analysis Intelligence Index v5.0 at 59 and scores 88.2% on SEC-bench Pro for finding vulnerabilities; WebDev Arena is up 89 Elo to 1769 (#4). It opens to cyber defenders this week, with wider access "coming soon." Introductory pricing is $2 per million input tokens and $10 per million output, rising to $4 and $20.

X user _mohansolo wrote that Carbon is performing well on the AGY (Antigravity) harness and that he's excited for people to try it (387 likes). A reply asked how soon, noting only cyber defenders can touch it. He answered that defenders need preparation and the right guardrails, and that Google wants to ship as soon as possible but can't take safety and security lightly.

Hands-on reports are mixed. dummerspast39 built a Minecraft-style game in about 30 minutes in one shot, ranking Carbon above Opus 5.5 and all GPT models but below Fable 5.5 (328 likes), though a reply said it looked like a decade-old app-store clone. Anca Dragan wrote that Carbon cut cheating and drew even fewer flags from a held-out misalignment monitor (129 likes). Developer theo criticized a model widely read as Argon for hallucinating in real codebases (916 likes); replies said a T3 Code integration bug fixed in 1.4.5 was to blame.

“Amazing, great to see this. Gemini 4 Carbon is a great model and performing well on the AGY harness. Excited for you all to try it soon!”

Next story →OpenAI lands $11 billion of a $30 billion target; Q2 revenue is about 60% of Anthropic'sRead an AI Insight →AI isn't replacing SREs, it's deskilling them: where does the intuition for the critical 5% come from?

Get tomorrow's stories as an alert

We'll ping you when each new edition or breaking story lands. Free, no sign-up.

Daily morning AI news, new AI Insights, and a monthly briefing of top stories on the 1st.

More stories from the same day

  1. OpenAI lands $11 billion of a $30 billion target; Q2 revenue is about 60% of Anthropic's
  2. ChatGPT and Codex usage limits merge in December, measured in tokens; instant chat excluded
  3. OpenAI's Altman makes surprise Korea trip, meets Samsung executives and gifts a "Codex micro keyboard"
  4. Musk: "Think of Grok Bot as a work colleague" as first stable version ships; usage guide draws 5,000 likes
See all of Sunday, November 1, 2026 →

* Summaries and publishing on this page are done by an automated agent. Check the facts against the original links.

Privacy PolicySupportSend feedbackClaude Code & CodexRSSListed on Sell With boost