If apple had a 1tb Mac studio, it’d sell so goddamn fast due to being able to run some of the largest open LLMs out there. I’ve been using GLM 5.2 and it’s damn near opus 5.6, the first model I thought was good enough to just write code without intense guidance. 1tb is well within the range where you could run the model on your own hardware. The higher prices are there to head off the deserved intense demand
Work bought me three 512gb Mac Studio M3 Ultras (one for my desktop and two for my lab). I run exo on two of them and can dang near run any model I want if I can find one that runs on the hardware (DeepSeek-v4-flash has been problematic so far). I’m so glad I got them before the prices skyrocketed. Hell I think they even stopped making the 512gb version completely.
If apple had a 1tb Mac studio, it’d sell so goddamn fast due to being able to run some of the largest open LLMs out there. I’ve been using GLM 5.2 and it’s damn near opus 5.6, the first model I thought was good enough to just write code without intense guidance. 1tb is well within the range where you could run the model on your own hardware. The higher prices are there to head off the deserved intense demand
Work bought me three 512gb Mac Studio M3 Ultras (one for my desktop and two for my lab). I run exo on two of them and can dang near run any model I want if I can find one that runs on the hardware (DeepSeek-v4-flash has been problematic so far). I’m so glad I got them before the prices skyrocketed. Hell I think they even stopped making the 512gb version completely.