The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Anyone running an open source LLM that's better than GPT / Gemini Pro?
| Dave Prole | 08/01/25 |
| JennStergerFan1488 | 08/01/25 |
| AP1488 8===D~~~ | 08/01/25 |
| Dave Prole | 08/01/25 |
| JennStergerFan1488 | 08/01/25 |
| Dave Prole | 08/01/25 |
| The Wandering Mercatores | 08/01/25 |
| JennStergerFan1488 | 08/01/25 |
| The Wandering Mercatores | 08/01/25 |
| Wang Hernandez | 08/01/25 |
| Theotokos is based | 08/01/25 |
| Mainlining the $ecret Truth of the Univer$e | 08/01/25 |
Poast new message in this thread
Date: August 1st, 2025 10:56 AM Author: Dave Prole (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148256) |
Date: August 1st, 2025 11:19 AM Author: JennStergerFan1488 you're not going to get better performance from an open source local model unless you've fine tuned it for something specific
the leading frontier models have better pretraining and also waaaay more compute
that being said the new open source models are so good that you can run them locally and get good enough performance. it's just slow |
 |
Date: August 1st, 2025 11:24 AM Author: AP1488 8===D~~~ Excellent poast/moniker synergy (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148337) |
|
Date: August 1st, 2025 11:33 AM Author: Dave Prole Which model do you recommended for shitlaw?
And which is best at making videos and creative content?
TYTYTY |
|
Date: August 1st, 2025 12:48 PM Author: JennStergerFan1488 literally just ask LLMs these questions
i'm not trying to be condescending that's actually how you should go about this. they will tell you exactly how to set up your own local system, based on what exactly you want |
|
Date: August 1st, 2025 8:18 PM Author: Dave Prole Which one should I ask? You seem more up to speed than all the ones I use (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49149895) |
|
Date: August 1st, 2025 11:23 AM Author: The Wandering Mercatores (from the Euphrates to the Forum) just wait for the open weights GPT to come out. its based on o4 mini and coming out later this month (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148333) |
|
Date: August 1st, 2025 11:26 AM Author: JennStergerFan1488 are they actually going to release that
i think they're just lying |
|
Date: August 1st, 2025 11:29 AM Author: The Wandering Mercatores (from the Euphrates to the Forum) yeah I have wondered that as well. they will probably say its not ready because of "safety" then release something gay in its place and delay it until further notice. honestly would be better if they released an open weights 4o that would be fire. o4 is obviously smarter but its reinforcement saftey training makes it no fun. will be good for making deranged bots though for committing cyber crimes (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148359) |
|
Date: August 1st, 2025 11:47 AM Author: Wang Hernandez who cares about this shit (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148399) |
|
Date: August 1st, 2025 12:43 PM Author: Theotokos is based Literally none of this matters and is irrelevant (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148587) |
|
Date: August 1st, 2025 1:34 PM Author: Mainlining the $ecret Truth of the Univer$e (You = Privy to The Great Becumming™ = Welcum to The Goodie Room™) My Mahchine™ (http://www.autoadmit.com/thread.php?thread_id=5757014&forum_id=2#49148778) |