undefined | Better HN

0 pointskarmasimida2y ago0 comments

He is all over the place, mixing tech specifics with unproven models.

Basically, he said he is happy with Mistral 8x7B and thinks it is on par/better comparing to OpenAI's closed source model.

0 comments

Well the lmsys leaderboard interestingly places it just 10 Elo under GPT 4 entries, so one would have to consider it somewhat on-par.

j / k navigate · click thread line to collapse

0 pointskarmasimida2y ago0 comments

He is all over the place, mixing tech specifics with unproven models.

Basically, he said he is happy with Mistral 8x7B and thinks it is on par/better comparing to OpenAI's closed source model.

Well the lmsys leaderboard interestingly places it just 10 Elo under GPT 4 entries, so one would have to consider it somewhat on-par.

j / k navigate · click thread line to collapse