News for hackers. Stuff that matters.

A comment on DeepSeek releases Janus Pro, a text-to-image generator [pdf]. Parent comment

Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by llm_trw on Monday January 27, 2025 @04:05PM (#42845616)
Llama is being gpt4 because they are releasing models 1/50th to 1/5th the size.

R1 is a 650b monster no one can run locally.

This is like complaining an electric bike inky goes up to 80km/h

[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by thot_experiment on Monday January 27, 2025 @06:44PM (#42847197)
R1 distills are still very very good. I've used Llama 405b and I would say dsr1-32b is about the same quality, or maybe a bit worse (subjectively within error) and the 70b distill is better.
[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by potamic on Tuesday January 28, 2025 @12:11AM (#42849053)
What hardware do you need to be able to run them?
[ Parent | on HN ]
Re:DeepSeek releases Janus Pro, a text-to-image generator [pdf] by llm_trw on Tuesday January 28, 2025 @02:21AM (#42849740)
The distils run on the same hardware as the llama models they are based on llama models anyway.

The full version... If you have to ask you can't afford it.

[ Parent | on HN ]