Hacker News .hnnew | past | comments | ask | show | jobs | submitlogin

The only point I am bit skeptical is that large language models are just not able to run on consumer hardware so the point with OpenAI is a bit moot.


The model runs on just about anything, it's just a question of how fast. My Intel i7 CPU can use OpenVINO to generate a standard image in about 25 minutes. The M1 can do it in ~45 seconds. For comparison, modern GPUs take about 10 seconds.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: