with this stuff they are just pushing me more towards 10k ai local pc.
and their subs are not even break even, meaning they are very cheap for the performance
how they want to ask for higher prices when they limit their own performance
Need two to run the glm/DeepSeek class of models. Even with a pro 6000 which is more expensive than the sparks the best you can run is qwen27b which is pretty mid. They also get faster the more you add to the cluster with tensor parallelism
Some guy vibecoded Adobe alternatives π.
Either adobe and other professional software companies hurry to make software for linux or they will be dead soon. Once the software compatibility is not an issue in linux people are going to leave in mass the microsoft shitshow