The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Why is AI so slow / why does it freeze so much?
| .,.,.;;,;.,;:,:,,:,.,:,::,..;.,:,.:;.:.,;.:.,:.::, | 01/12/26 |
| screenman | 01/12/26 |
| fully online and responsive | 01/12/26 |
| chopped unc | 01/12/26 |
| Jared Baumeister | 01/12/26 |
| chopped unc | 01/12/26 |
| Jared Baumeister | 01/12/26 |
Poast new message in this thread
Date: January 12th, 2026 5:47 PM Author: .,.,.;;,;.,;:,:,,:,.,:,::,..;.,:,.:;.:.,;.:.,:.::, (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584226) |
Date: January 12th, 2026 5:47 PM Author: screenman you're not paying $2k a month for it yet (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584229) |
|
Date: January 12th, 2026 5:50 PM Author: fully online and responsive Each query has to kill an endangered species (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584234) |
|
Date: January 12th, 2026 5:51 PM Author: chopped unc the apis are fine if you use those. if you are using the web app or desktop app there are a lot of things that can make it go slow. having a ttt computer being one of the main reasons. check your cpu usage when its streaming and if it spikes up to like 100% with a ton of threads then you just need to open a new session. the app keeps the entire conversation history in memory and on screen and as the session grows the interface has to maintain and update every previous messsage. guarantee you its all UI/render latency not model inference latency. for instance claude streams fast as fuck on amazon bedrock like almost instant. (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584242) |
|
Date: January 12th, 2026 5:54 PM Author: Jared Baumeister (mail@baumeisterlaw.com) because there's a shortage of GPUs and RAM. Most OpenAI users are only interacting with quantized models because why let proles use the full uncompressed model?
https://www.xoxohth.com/thread.php?thread_id=5820268&forum_id=2 |
 |
Date: January 12th, 2026 5:58 PM Author: chopped unc no. all of the "slow ai" issues are always slow streaming which is a ui rendering issue caused by not a powerful enough local cpu and buggy client optimization (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584255) |
|
Date: January 12th, 2026 6:06 PM Author: Jared Baumeister (mail@baumeisterlaw.com) makes sense (http://www.autoadmit.com/thread.php?thread_id=5821222&forum_id=2.#49584270) |