AI Judge changed title after evaluation, original title: "OpenAI runs simultaneous A/B tests on updated text and image models inside ChatGPT"
OpenAI often launches models after major Google announcements.
I love GPT-5.5. It's a workhorse and exactly the model I was hoping for. But the fact that rumors say version 5.6 is already in the starting blocks makes me even more excited!
OpenAI is on fire.
🚨 ChatGPT is running TWO A/B tests at once right now:
→ text model → image model both showed up yesterday and today.
can smell 5.6 from a distance 🫣
Many users praised GPT-5.5 as a reliable workhorse and celebrated OpenAI's rapid upgrade pace amid GPT-5.6 rumors, while others criticized recent performance drops and overly frequent releases.
No Digg Deeper questions have been answered for this story yet.
I love GPT-5.5. It's a workhorse and exactly the model I was hoping for. But the fact that rumors say version 5.6 is already in the starting blocks makes me even more excited!
OpenAI is on fire.
Big week for AI
Google is dropping 3.2 and it’s a pretty good model
Historically OpenAI has countered with their own launch - GPT 5.6 may land as well
@kimmonismus I'm more waiting for Codex computer use on windows and in EU than GPT-5.6 😅
@kimmonismus You could choose better images to show comparison 5.6 vs 5.5...
@kimmonismus They lie about available features and have no customer support of any kind. Fuck them
@bindureddy The frontier model cold war is the most predictable cycle in tech, but at least it means cheaper inference for the rest of us.
@rafal_darlak @kimmonismus No in the second Picture is 5.6 vs 5.5 and in the first Imagenv2 vs v2.1
@kimmonismus GPT-5.5 feels like the AI equivalent of a really reliable Toyota. And that’s a huge compliment.🚗🤖😁
@kimmonismus after about 2 weeks with a new model, i’m like ‘alright, what’s next?’
they seem to be compressing release schedules, so it might actually come this month
@kimmonismus IDK why but GPT seems to work totally different in codex environment. It seems to behave as just another model. @romainhuet did some real magic
@kimmonismus Once you try gpt 5.5 you don’t go back
@kimmonismus Are we sure that 5.6 will be better? Sometimes it's just worse in a way or in a lot.
@kimmonismus swap providers most weeks tbh, version churn matters less than people think
@kimmonismus Workhorse is a better compliment than flashy benchmark numbers. Release cadence is starting to make planning feel pointless.
@kimmonismus yup but also quite expensive tbh. one could argue it is worth it because if say we used kimi k2.6, with 40 usd a month we get more tokens but worse performance but anyways my 20 bucks codex hits limits within 30 minute max when using in medium thinking with 5.4
@kimmonismus can't even finish onboarding to one model before the next one drops lmao. but yeah 5.5 genuinely slaps, feels like a proper daily driver
@kimmonismus honest take: "workhorse" just means it doesn't break when you ask it something normal. we really lowered the bar somewhere along the way. solid model though, no complaints
@kimmonismus “5.6 in the starting blocks” is the tell.
you don’t even finish wiring 5.5 into prod, evals still settling, and the cadence is already lapping you.
sota has an expiration date now.
just keep shipping.
@kimmonismus GPT-5.6 better not be another “we shrunk the brain to save money” release
Exciting week but it also widens the gap. As cloud frontier models get heavier and more expensive for high stakes reasoning, the need for a robust local/cheap API routing tier becomes critical. You can't run a scalable agent workflow if every single micro task pays the premium 'frontier tax'