GPT-6 Astra Customers Say OpenAI's Latest Mannequin Bought Dumber. It Occurred Earlier than, Too – Decrypt

Briefly
Astra “nerf” complaints are flooding X only a week after launch.
Beforehand, customers discovered the writing high quality difficulty existed at launch.
OpenAI's final mannequin, Sol, noticed the identical backlash in July.
Every week in the past, GPT-6 Astra was rebuilding Manhattan road by road inside a recreation engine, wowing customers left and proper with its capabilities. This week, the identical crowd is posting screenshots asking OpenAI what occurred to their mannequin.”Astra feels considerably dumber for me at present,” the favored pseudonymous developer synthwavedd posted on X. “Was solely a matter of time earlier than The Put up-Launch Lobotomy. Disgrace.”Myriad: How excessive will Nvidia inventory go? Click on to make your prediction.It is a acquainted chorus, with AI customers accustomed to their favourite chatbots and brokers out of the blue feeling like they have been “nerfed.” More often than not there are cheap explanations for the perceived drop in efficiency from these fashions. However when sufficient individuals begin saying the identical factor on the similar time, it is value inspecting what is going on on. And social media in the meanwhile is replete with examples of customers shortly souring on GPT-6 Astra.Developer Pranjal Paliwal, who beforehand had praised the mannequin, went again and truly learn the code Astra had written him. He didn't like what he discovered.”We do not have AGI,” he posted. “Now we have a regression.”
fml, lastly had a have a look at the code astra wrote. I take my phrases again.
We do not have AGI. Now we have a regression.
How can or not it's so sensible and so dumb on the similar time!
— Pranjal Paliwal (@betterclever) September 11, 2026AGI, brief for synthetic basic intelligence, is the trade's time period for a machine that may do basically something a human can do cognitively. OpenAI's personal president used the phrase at Astra's launch. Every week later, certainly one of his customers is utilizing it to explain the alternative.The complaints share the same form. Developer Pankaj Kumar listed the signs: quicker solutions, worse high quality, and a suspicion OpenAI “diminished the juice worth.” Founder Saba requested OpenAI straight why she now has to “dumb it down” to get duties performed.”Juice worth” is not an official time period. No one has outlined it. Nevertheless it's shorthand everybody right here understands: the quantity of computing effort the mannequin spends pondering earlier than it solutions, and the suspicion that OpenAI quietly turned that dial down as soon as the launch demos had performed their job.Some individuals examined it correctly. Salio and researcher Md Ismail Sojal each ran the equivalent immediate in opposition to launch-day Astra and at present's model, and each acquired worse outcomes from the present one.
In the present day’s GPT-6 Astra output seems to be worse.GPT-6 Astra at launch vs GPT-6 Astra at present.
– Tibo stated they've compute – Similar immediate,Similar settings.- evan I ran the very same immediate on GPT-6 Astra at launch and once more at present. – The distinction is greater than I anticipated.…
— Md Ismail Šojal 🕷️ (@0x0SojalSec) September 11, 2026Others are simply switching again. Dax Raad, who builds the coding instrument Opencode, stated his staff has gone again to Astra's predecessor, GPT-5.6 Sol, as a result of the spend doubled for downsides that weren't value it. ChatGPT consumer Mustafa Sahinli put it extra bluntly: Astra now makes him really feel like Claude Opus 4.6 “after 1 week of launch,” a jab at Anthropic's personal post-launch backlash.Not everybody thinks OpenAI has purposely nerfed its latest mannequin although. The pseudonymous consumer Antikythera revealed essentially the most detailed rebuttal within the pile, arguing the timeline runs backwards.”It's as dumb because it was on launch,” wrote Antikythera. “The mannequin is sweet, however the mannequin has numerous issues. It is lazy. Writes like a bullet-point-addict… individuals have been overhyped on launch week, now that they had time to check it and see its errors.”In different phrases: nothing modified. You simply stopped being dazzled lengthy sufficient to note the mannequin was all the time slightly little bit of a bullet-point-addict.T3Chat founder, Theo, has a considerably related concept: Astra is definitely extra inconsistent than Claude Fable, so it could produce excellent or downright silly code at occasions. What appears to be taking place is a wave of customers posting the dumb outcomes extra regularly now that the honeymoon is over.
GPT-6 Astra has performed unbelievable issues I by no means thought a mannequin might do. It has additionally performed among the stupidest issues I've ever seen a mannequin do.
Typically talking, Fable 5.1 simply does what I ask. pic.twitter.com/X2UwI7tDnD
— Theo – t3.gg (@theo) September 8, 2026This is not new, both. OpenAI's final flagship, GPT-5.6 Sol, went by way of the equivalent cycle in July, when customers reported its prime reasoning mode had gone shallow in a single day. OpenAI govt Tibo Sottiaux denied intentionally weakening it, whereas confirming the corporate had been experimenting with reasoning effort, the setting that controls what number of steps a mannequin “thinks” by way of earlier than answering.One reply summed up the operating joke: closed labs launch a mannequin, and it catches “some type of illness a number of days later and out of the blue turn into[s] dumber.” One other supplied the cynic's idea in a single line: “They in all probability get quantized so they don't seem to be burning the businesses as a lot cash.”
this time they have been dumb on arrival although. I gave Astra precisely one activity, which it failed miserably, 5.6 SOL on xHigh then did the cleanup…
— Physicist 🏖️ 💻 (@heisenberg_btc) September 11, 2026Quantizing a mannequin means shrinking the precision of its inside math to chop prices, typically at some expense to accuracy. OpenAI has by no means confirmed doing that on objective to a shipped mannequin.OpenAI has not issued a Sol-style assertion about Astra but. The mannequin stays the corporate's first to cross what it calls the crucial threshold for cybersecurity threat, which means it could discover and chain collectively unknown software program vulnerabilities with out human steerage, a functionality restricted to vetted defenders beneath OpenAI's Dawn program.Dumber or not, it nonetheless prices $10 per million enter tokens and $50 per million output tokens, 2.5 occasions what Sol charged at launch.Every day Debrief NewsletterStart daily with the highest information tales proper now, plus authentic options, a podcast, movies and extra.