Deepseek v4.1 flash vs secret model
Tested both with the same prompt at the highest reasoning available v4.1 flash took 2 hours to finish both tests and cost $2.60 secret model finished both in 30 minutes, can’t reveal cost yet v4.1 flash spent a lot of time overthinking
Nathan is right about the open ecosystem being underrated.
But the reason is bigger than "loads of new fun models." I just watched Stanford CS336 Language Models from Scratch. Percy Liang teaches it. On his slide about the state of the field, he highlights one line for the
Which model do you think made this?
Used gauntlet loop and it one-shotted this basic game around an hour to build less than $5 way faster and better than the last time i tested it
New Deepseek 4.1 Flash model.
I'm seeing MUCH better results than the previous Flash. This is pretty great for a flash model.
You have reached the end of the archive
All of deepseek