Fara 7B: A cheap & capable open weights computer use agent (CuA) they got within a few points of o3’s performance using only 4k training data points (yes, synthetic) www.microsoft.com/en-us/resear...

community note: using cost on the y axis makes it appear like cheaper models are more capable on pass@3

Follow

@timkellogg.me yeah, if the first attempt fails by wiring money to the wrong account, I don't care that it didn't cost as much in compute. They then omit the better agent / models in the table where they only compare accuracy...

Sign in to participate in the conversation
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.