I ran through my first inference model, DeepSeek-R1. The free GPU obtained by registering on kaggle can be used for 30 hours per week for free.
Notebook:
This changes how you cheap out on ai forever
Moon dev: "i measured it on a real 40 line brief, three runs each, and holy hell what came back stopped me cold" "deepseek took 50 seconds and then glm flash did the exact same thing in 10 and i just stared at the screen" "i applied
Life at Veritylab after using the last DeepSeek after OpenAI dowgraded us to the 100$ tier
Back to testing DeepSeek-V4.1-Flash more
Cause this model is too good. One-shot god model. I'm really loving these infinite world's it's able to spin up in one-shot, so decided to take it a step further, i asked it to build a infinite city: Sapporo in heavy snow at blue hour.
You have reached the end of the archive
All of deepseek