Good news! The actual cost of using AI is going through the fucking roof
| Balding brunch | 05/24/26 | | Khaki stimulating tank theater stage | 05/24/26 | | Sepia water buffalo | 05/24/26 | | Stirring spot | 05/25/26 | | Balding brunch | 05/25/26 | | sienna halford | 05/25/26 | | demanding faggotry whorehouse | 05/30/26 | | Odious mildly autistic legend | 05/30/26 | | demanding faggotry whorehouse | 05/30/26 | | bossy multi-billionaire | 05/31/26 | | Sepia water buffalo | 05/30/26 | | Odious mildly autistic legend | 05/30/26 | | Sepia water buffalo | 05/30/26 | | Microsoft Execution Containers | 06/04/26 | | Goyim in Abundance | 06/04/26 | | ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... | 06/04/26 | | Microsoft Execution Containers | 06/04/26 | | POPE LEO IS WEAK ON CRIME | 06/04/26 | | AZNgirl Covered in Lox Schmear for Lindsay Graham | 06/04/26 | | Microsoft Execution Containers | 06/04/26 | | Emotionally + Physically Abusive Ex-Husband | 06/04/26 | | LathamTouchedMe | 06/04/26 |
Poast new message in this thread
Date: June 4th, 2026 7:28 PM
Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,...
The costs at a fixed level of intelligence are declining rapidly. Qwen 3.6 is dirt cheap, runs on consumer hardware and is better than a frontier model from a year ago. The costs of frontier models are going up because people want to burn a shitload of tokens with the absolutely best model (which are increasingly large), but whether that makes sense is another matter. In most case for agentic workflows, using a frontier model to generate a plan and have a lower end model follow it makes more sense
(http://www.autoadmit.com/thread.php?thread_id=5868889&forum_id=2)#49916039) |
|
|