Gemini 3.5 Pro is so bad like even with multiple iteration it can't make a blackhole
42K followers
52,286 views
36 Comments
Comments from the original post on X (Twitter)

@jan_volad
Imagine someone from the GPT-3/3.5 era (2020-2022) finding out that in 2026 people are disappointed because a model can’t make a fully working 3D black hole visualization 😭
23 likes

@Dreip8
I think the revolution will come with Gemini 4, apparently it has to change the entire pre-training data.
11 likes

@SpecjalistaMSS
And now you'll get a weak model in an environment that works against it like Antigravity or gemini-app, so the results will be even worse. And damn it, I believed in them and bought a one-year AI PRO plan. I feel angry.
8 likes

@Vicrom1509
And the release of "this" is today?
3 likes

@Gdgtify
I was very high on Gemini but we got to see some major move at some point surely.
2 likes

@MatthewSoroVC
Google is slow and burdened by 10+ years of non-meritocratic hiring. Turns out when you’re hiring to fill race and gender quotas, you no longer have the best. Well deserved. Fuck off, Google
2 likes

@abhinavflac
Bro gemini feels like its chasing the frontier instead of defining it
2 likes

@SimonasLTU1
Such a shame to see this, I hope Gemini 4 will get them back on the race..
1 like

@Sora_Kun_94
Are we sure this isn't low reasoning or any weird stuff happening behind? This result seems worse than 3.1
1 like

@notjazii
muse spark can do better lmaoo
1 like

@chetaslua
Yes
3 likes

@pingouinism
Bro, stop or they will postpone the release
1 like

@BelethDionysian
@GeminiApp Aren't you embarrassed? You have access to arguably the largest pool of data and resources on the planet, yet you still managed to build such a painfully mediocre LLM. It's honestly impressive how much potential Google managed to waste with Gemini.

@tolgaozisik
🥲

@Robertwithai
I asked to make a website recently. It was the same ai slop design other ais had 1 year ago. Non-sense.

@realunclect
🧢🧢🧢

@chetaslua
lol test it arena yourself

@sebuzdugan
i get cleaner gemini outputs by setting a strict feature budget before asking for visuals

@TLMagnus
Gemini may be developing, but it seems like it's taking precautions against something.

@Blazerdan142102
How many points do you expectate it to have in artificial analisys index? 55?















