The most prestigious law school admissions discussion board in the world.
|
 |
 
|
I can do shit with Qwen3.5 that not even Chat-GPT can do
| wine foreskin | 02/26/26 |
| ruby excitant fortuitous meteor stain | 02/27/26 |
| wine foreskin | 02/27/26 |
| ruby excitant fortuitous meteor stain | 02/27/26 |
| wine foreskin | 02/27/26 |
| awkward chapel | 02/27/26 |
| wine foreskin | 02/27/26 |
Poast new message in this thread
Date: February 26th, 2026 11:48 PM Author: wine foreskin I checked with Claude and asked if this was possible
https://i.imgur.com/yAOtRAk.png |
Date: February 27th, 2026 12:01 AM Author: ruby excitant fortuitous meteor stain None of those gpt models exist anymore (http://www.autoadmit.com/thread.php?thread_id=5838842&forum_id=2...id.#49698781) |
 |
Date: February 27th, 2026 12:11 AM Author: wine foreskin Whatever man, I'm pretty sure OpenAI won't just spot you 28gb of VRAM on demand (http://www.autoadmit.com/thread.php?thread_id=5838842&forum_id=2...id.#49698794) |
|
Date: February 27th, 2026 12:39 AM Author: ruby excitant fortuitous meteor stain Show me qwen3.5 do reasoning as good as gpt 4.5/o3 and I'll be impressed
Anybody can buy higher context limits with a local llm. It's a cost function, not a major technical hurdle like higher level reasoning |
|
Date: February 27th, 2026 1:16 AM Author: wine foreskin I'm not saying Qwen wins at everything. And even if a question of resource constraints, it's still true that ChatGPT won't give me the same context window size. I don't even know what the upper limit is on Qwen. Pretty sure I can make it 3-4x what it is now without losing any quality. Deepseek 4 is supposed to have a 1mm token context window (http://www.autoadmit.com/thread.php?thread_id=5838842&forum_id=2...id.#49698850) |
|
Date: February 27th, 2026 12:07 AM Author: awkward chapel You're assuming the tokens are equal among models. They aren't (http://www.autoadmit.com/thread.php?thread_id=5838842&forum_id=2...id.#49698790) |
|
Date: February 27th, 2026 12:09 AM Author: wine foreskin I didn't say shit about tokens, Claude did. (http://www.autoadmit.com/thread.php?thread_id=5838842&forum_id=2...id.#49698791) |