listening now
Xi makes a speech that's basically DeepSeek recruiting copy. AI as steam and electricity. «We human beings must answer the question posed by our times. How to get along with thinking machines?». Open source AI for all humanity is part of Xi's answer.
Buckle up.
Whoa, they buried the most interesting part of all of this: Inkling-Small: 276B A12B, coming soon.
Multi-modal at that size w/ those scores if they hold thru til release is very strong (see their blog for the full list of results)
Whoa, they buried the most interesting part of all of this: Inkling-Small: 276B A12B, coming soon.
Multi-modal at that size w/ those scores if they hold thru til release is very strong (see their blog for the full list of results)
@Hesamation If the model that's infamously 10x cheaper than everything else has 70-80% profit margins... the average discourse around API rates being "subsidized" is way off
https://t.co/OCZNpe0yp5
> DeepSeek's gross margin from selling v4 API access is...70% to 80% (!)
I wonder if Wenfeng would still say «not to subsidize, nor to reap excessive profits. This price also has a slight profit margin on top of cost»
I think so. That's "slight".
Dario is one hell of a humanist.
I never considered that "room temperature IQ" refers to fahrenheit, I thought the point was that someone is a braindead slug creature with an IQ of 20, which is so much better
@Thesmophilist The armed underground resistance faction planning out their first big move to bring about the revolution: "so we're all agreed that Anne Widdecomb is simply too powerful"
@ivanfioravanti I'm getting ~178 tok/s prefill and ~10.5 tok/s decode on a single Spark. Currently it requires a patched version of llama.cpp that's pinned to an old commit form over a month ago, I imagine there's room for improvement.