Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon! It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback. Here’s a look at the benchmarks:

Sep 30, 2026 · 8:02 PM UTC

1,372
2,330
25,905
3,723,937
Importantly Argon has frontier safeguards and we are rolling it out responsibly - it’s with the US gov’t and going to a set of trusted cyber defenders through our Fairwind Program today. We’re going to make it available as soon as we can and as safely as we can. So hold tight, lots more coming, and you’re going to see us iterating rapidly. blog.google/innovation-and-a…
112
113
2,424
277,670
Sort replies: Relevant Recent Liked
Replying to @sundarpichai
How do I try it? Does Google have a Claude Code equivalent?
13
5
31,899
Replying to @sundarpichai
@sundarpichai desperate attempt to stay relevant. Release the model or fuck off
1
660
Replying to @sundarpichai
5
29
952
54,554
Replying to @sundarpichai
Gemini 4 wiping out the industry
13
20
376
47,035
Replying to @sundarpichai
Well if any Indi gamers on here start displaying its performance then we will truly know how well the model works.
1
1,567
Replying to @sundarpichai
Facts, from a Google fan boy (well not exactly). If you do not have consistent flow of good models and Google is left behind for 10 of the 12 months of the year, and antigravity continues to be the worst harness out there, it will not matter
1
55
9,027
Replying to @sundarpichai
@sundarpichai : Please invest in making a great code editor like Cursor or VS Code, where we can access Gemini models. Benchmarks are great, but if it's difficult to use Gemini models, it's of no use to the general user. Or make them generally available at Cursor, etc.
10
21
18,156
Replying to @sundarpichai
benchmaxxed
7
11
627
23,293
Replying to @sundarpichai
what are the chances this is benchmaxed
2
1
13,971
Replying to @sundarpichai
pricing
2
20
4,566
Replying to @sundarpichai
just that in real world scenarios Gemini is not delivering. the back and forth and the mistakes are just so frustrating.
2
10
14,866
Replying to @sundarpichai
solid
7
7
393
26,910
Replying to @sundarpichai
I never stopped my subscription because I'm expecting something very fast and powerful...
2
2
8,869
Replying to @sundarpichai
Constantly switching between frontier models 😅
2
4,014
Replying to @sundarpichai
how do security researchers get access as trusted defenders?
2
2
11,761
Replying to @sundarpichai
Why should we pay for it?
1,958
Replying to @sundarpichai
"What ever happened to Gemini?"
9
4,731
Replying to @sundarpichai
1
4
2,896
Replying to @sundarpichai
i'm not sure i would have released this while still behind on some software engineering benchmarks.
2
1
6,076
Replying to @sundarpichai
Chad Sundar
2
3
74
7,456
Replying to @sundarpichai
2
2
22
2,955
Replying to @sundarpichai
1
1
10
1,617
Replying to @sundarpichai
Finally, a good name. Argon.
4
57
16,209
Replying to @sundarpichai
They proved us wrong 😭
1
132
13,186
Replying to @sundarpichai
you released the same image with previous gemini models and we all know how bad those turned out to be... until normal devs get access to use the model, this means nothing.
1
1,451
Replying to @sundarpichai
That's a great, you just forgot to ship it. Maybe you can't ship it without nerfing it, since you guys don't have compute.
3
6
3,551
Replying to @sundarpichai
They renamed it so that we wouldn't shit on them as much... Won't work.
28
13,091
Replying to @sundarpichai
Bro, can you please only announce when you can actually give access to everyone? Like, why tf do I care about Cyber Defenders and shit?
1
8,828
Replying to @sundarpichai
Tried many times earlier hoping it to get good output but always disappointed , hope this time it turns well and feel worth paying for the plan
2
2,562
Replying to @sundarpichai
1
4
3,771
Replying to @sundarpichai
Well, I don't really trust DeepSWE much these days. It does seem to perform decently on the other benches too, but hopefully it wasn't just benchmaxxed. My concern as always with Gemini models is the hallucination.. Hope this one isn't smoking something weird.
5
7,444
Replying to @sundarpichai
Excited to test and ship it in @CommandCodeAI
1
27
4,645
Replying to @sundarpichai
google do something 🥲
2
1
97
7,178
Replying to @sundarpichai
By the time this is available for public we will see opus 6 and the cycle continues
1
367