Google DeepMind launches Gemini 4 Carbon, SOTA on cyber and long-horizon tasks; not publicly available
Google DeepMind released Gemini 4 Carbon, its next frontier model, claiming SOTA on cybersecurity and long-horizon tasks, with initial access limited to cyber defenders to iterate on guardrails. Nathan Lambert criticized the announcement, saying the model isn't publicly usable and score announcements without access are a miss; he later speculated the timing was tied to earnings and softened his tone. Lambert also suggested the release is aimed at pre-empting a competing Fable 5.5 launch.
All posts · 31, most important first
- Carbon can work through complex tasks for days without getting stuck or taking shortcuts – delivering high-quality, reliable code. Feedback from an initial group of cyber defenders will help us iterate on guardrails before we expand access soon. Find out more ↓ goo.gle/4BCFvj8120 likes · 3 reposts
- Gdmi whenever I want to try a Gemini model it’s not publicly available. Don’t bother announcing your scores if we can’t use it.
7 likes · 1 reposts - This Gemini 4 Carbon model getting teased today and not publicly available seems like a graspy/last ditch move to sneak it in before Fable 5.5 blows their socks off.147 likes · 3 reposts
- the fact that its for earnings just is sus, but other labs deserve some frustration here too5 likes · 0 reposts
- Gemini 4 Carbon is our next frontier model. It’s state-of-the-art across cybersecurity and long-horizon capabilities.
1,424 likes · 342 reposts - Honestly I thought you learned one of the last few times, this is just total swing and miss. Rookie move.0 likes · 0 reposts
- ⚡Iterating at lightning speed! Gemini 4 Carbon is SOTA across key benchmarks and combines frontier capabilities with deep safety & security research.
1,061 likes · 109 reposts - Gemini 4 Carbon benchmarks;
206 likes · 13 reposts - Behold the new king.
92 likes · 12 reposts - More info in the blog: blog.google/innovation-and…40 likes · 1 reposts
- Okay, Google, next time just wait the extra two weeks, what is this nonsense, do you know how much time it takes to read one of those safety cards and do a proper eval set?I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model. Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards. As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight! open on X29 likes · 1 reposts
- 22 likes · 0 reposts
- I feel a little bad ranting at Gemini about the unreleased model, as going through the unofficial channels to get a frontier model out today I hear is a good bit of a mess and definitely not easy. That's why they say patience is a virtue.14 likes · 2 reposts
- I honestly don’t know how to try Gemini in the agentic era, maybe I finally need to fix that. Pretty wild they just have been out of it since December 2025 but look maybe back.I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model. Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards. As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight! open on X12 likes · 0 reposts
- Sir Demis:@demishassabis: ⚡Iterating at lightning speed! Gemini 4 Carbon is SOTA across key benchmarks and combines frontier capabilities with deep safety & security research. open on X11 likes · 0 reposts
- It's happening again.10 likes · 1 reposts
- Sundar:I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model. Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards. As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight! open on X10 likes · 0 reposts
- Logan:Introducing Gemini 4 Carbon, it's a big jump over Argon which we released just 2 weeks ago. Gemini 4 Carbon is SOTA on cyber, long horizon tasks, ranks #1 on Artificial Analysis V5, significantly improves on usability in domains like coding, and is more aligned. open on X10 likes · 0 reposts
- Okay, so the explanation for 'why does Google keep saying they are doing RSI when their safety card says they definitely aren't doing RSI' is that no one at Google was ever talking about Argon? Must... control... fist... of... doom...8 likes · 0 reposts
- This is where I change tomorrow's section 'Gemini 4 Argon Is a Good Model But Not Fully Frontier' by changing 'Is' to 'Was'...@andrewcurran_: Gemini 4 Carbon benchmarks; open on X6 likes · 0 reposts
Gemini 4 Carbon is #2 on the Agent Arena, right behind Opus 5.5 (Max). It's #4 on web development capabilities, and ranked #1 on text-based conversations. It pushes the pareto frontier, ~4x cheaper than Opus 5.5. By any measure, Google is back on the frontier! open on X6 likes · 0 reposts- How do I put this: Safety card. Now. Like, now.5 likes · 0 reposts
- All on the leadership not on the ICs at all. ic’s clearly cooked hard.5 likes · 0 reposts
- 'We've been dogfooding continuously for weeks, and we've almost adapted to the new intelligence -- taking ambitious code rewrites for granted, leaving loops overnight that launch dozens of experiments to repro, root-cause, and fix mysterious bugs, and using swarms to inspect and fix RL data.'3 likes · 0 reposts
- Things move even faster lately.3 likes · 0 reposts
- Krypton Carbon Argon Flash2 likes · 0 reposts
- @natolambert: I feel a little bad ranting at Gemini about the unreleased model, as going through the unofficial channels to get a frontier model out today I hear is a good bit of a mess and definitely not easy. That's why they say patience is a virtue. open on X1 likes · 0 reposts
- Logan says it's #1.1 likes · 0 reposts
- @natolambert: I feel a little bad ranting at Gemini about the unreleased model, as going through the unofficial channels to get a frontier model out today I hear is a good bit of a mess and definitely not easy. That's why they say patience is a virtue. open on X0 likes · 0 reposts
- @natolambert: I feel a little bad ranting at Gemini about the unreleased model, as going through the unofficial channels to get a frontier model out today I hear is a good bit of a mess and definitely not easy. That's why they say patience is a virtue. open on X0 likes · 0 reposts
- Oh I actually CANT. Lmfao@natolambert: Gdmi whenever I want to try a Gemini model it’s not publicly available. Don’t bother announcing your scores if we can’t use it. open on X0 likes · 0 reposts
@googledeepmind org · 
@natolambert research · 
@demishassabis lab · 
@andrewcurran_ commentary · 
@thezvi commentary · 