I found the old Gemini 3.1 Pro SVG demo from February and was curious to see how it would compare to GPT-6 Astra 7 months of progress

Thank you OpenAI! :)

So many cool game demos with Opus 5.5 I really recommend following @JaydenDavisNC, his demos always impress me

ChatGPT Pro Max? I don't have enough money to test that 😭

Why is the goal getting a Fable/GPT-6 level model in 2 to 3 months when we'll have much better models by that time? This sounds like they gave up on being a frontier lab

Check out this scene that M1 made with GPT-6 Astra Genuinely insane

Love to see how Meta and Anthropic employees are much more active here lately

I can get much better results out of GPT-6 Astra but it needs a lot of hand-holding, Opus 5.5 has better taste out of the box But ngl this one is pretty cool

Whatever makes OpenAI models add so much text needs to be fixed, seriously, why did it add this?

Opus 5.5 vs GPT-6 Astra A solar system in 512x512 pixel art Why does Astra love adding text so much? I love Opus' result, it's much cleaner

It really annoys me when people say "don't trust benchmarks, try the models yourself" to make excuses for a model not being good It's one thing to question benchmarks when the results are mixed, it's another when 99% of them show zero improvement or even a regression in capabilities Let's be serious

No one will ever convince me that this isn't art

"A beautiful fountain" Opus 5.5 vs GPT-6 Astra Both on max reasoning effort

I would say Opus 5.5 is as good in Minecraft as GPT-6 Astra while having better taste Can't wait to see how it performs in VoxelBench

This is really cool Made with Opus 5.5

I get that this release was about efficiency and I'm very happy about that but why did they have to call them GPT-6 instead of keeping the 5.6 name

GPT-6 Sol vs Opus 5.5 As they say, a picture is worth a thousand words...

And yeah, I know I might be hating a bit too much again, but I don't care when other companies do this because I DON'T use their products I like OpenAI, I like their products, I prefer Codex to Claude Code, and it sucks knowing I could get something much better by paying for Claude because I don't want to subscribe to Claude So yeah, when an OpenAI model doesn't live up to expectations I'll speak up and complain I don't think there's anything wrong with that and they usually take feedback pretty well

This is exactly why you should never release your top-tier model before your mid-tier and low-tier ones...